JP2594641C - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION   The present invention guarantees privacy within a data network, and more particularly within this network
For the protocol for.   With a large number of computers, with a large amount of distributed computing,
Includes personal computers, workstations, and databases
In a data processing system, a large amount of data is exchanged between these data processing systems.
It is frequently required to change the data. These exchanges require a communications network. This
Communications networks exceed the geographic range of a local area network, but extend
Used to interconnect data processing systems in smaller areas than enclosures
Is called a metropolitan area network, where high-speed data traffic is
Requires a data network capable of transmitting in time.   As the requirements for data communication increase, the use of carrier data networks will increase.
More and more attractive. This network is shared by multiple users. With this, high-speed data communication equipment such as optical fiber networks can be shared
It works. These networks require an effective high degree of security.   In a prior art carrier data network, the user is usually
Access to the forwarding mechanism is obtained using a suitable password structure. The user
With access to the data network, fees for using the network are allocated to these users.
Users who are not authorized to use this carrier network cannot access the network.
I will be. Then, the user can use another terminal, for example, a database system.
Once the data path to the system or computer mainframe is obtained, additional paths
The password configuration authenticates this user and the database system or main
Frame is not accessed by unauthorized users of this system.
Used to keep. However, the user cannot decrypt through the user's password.
Once access to the data network is obtained, a further check of this user's identity is performed.
It will not be done. Security data network One of the most frequently used tactics to penetrate illegally is to allow illegal users
Pretending to be user attributes.   In a shared media carrier network, such as a local area network, packet
Checking the source means that individual users have direct access to this shared medium.
Especially difficult to have.   One problem of the prior art is therefore that within a carrier network the access of individual users of the
Access capabilities to be consistently authenticated after a user logs into this system.
There is no efficient device. Unauthorized users are allowed and logged in
Access of the user by prefixing this packet with his identity.
It is possible to capture.   Protocols for data transmission systems, e.g. 25 steps
The protocol calls these blocks in a flexible way in a limited length block.
Provide equipment for transmission between end users connected to the system. X. 25 is many
No A variable number of resources for connections between specific end users, including
It is prepared for transmitting data over a network requesting a link.   X. Protocols such as 25 mainly use basic data transmission messages in individual blocks.
Transmits relatively small amounts of data, typically limited to thousands of bytes
It is intended to be used in such systems. As a result, these systems
The length of the header information transmitted with each block is considered for efficiency.
Is limited to information that is absolutely necessary for data transmission processing.   However, very large amounts of data, e.g. all multiple files, are frequently transmitted.
For systems such as X. The 25 protocol is not very efficient. No.
1, X. The 25 protocol adds a significant amount of error checking to the network (layer 3 functionality).
) Is performed; secondly, if an error is detected, or
If no is received in the network, a transmission retry is automatically performed; and
3 shows a series of data messages Timing is performed to ensure that locks are sent within a predetermined time period.
You. X. In the 25-data transmission protocol, most messages are
Transmitted after the connection between the end users is established in the network. Between these end users
The process of establishing a connection is a routine for transmitting data between end users.
Set up the buffering table and allocate buffer resources for these messages
Including the operation. The tendency to use significant amounts of network resources exclusively for individual connections
This configuration allows the transmission of a large number of data transactions between end users.
If it is adopted, it causes a big problem. In addition, individual end users
Due to temporal relationships, the number of resources to be allocated in the network in advance cannot be controlled.
It becomes huge.   Typically, today's X.R. 25 networks are prepared for very high error rates,
Check data frequently to ensure that certain blocks are error free.
Do. In addition, these networks have an alternate routing capability and a certain data
Bro Block is transmitted on a shorter path than other blocks, and
May reach first. This is especially true in metropolitan area networks.
In very rare cases, check this for individual data blocks.
A significant amount of processing resources is allocated to the task.   In addition, X. The 25 protocol strictly checks for errors, but adds
Has room for improvement. Specifically, an illegal user logs in to the system first.
Can pretend to be a user with permission.   Large local area networks are limited to small metropolitan area networks.
Has a defined range of applications. However, these local area networks are large.
Lack of capacity to serve a population of
Yes, and if they are used as a carrier network, their integrity is also inadequate
.   One problem of the prior art is therefore the protocol used to transmit the data.
Unauthorized call Do not provide sufficient protection to prevent unauthorized access by users.
And In addition, these protocols carry numerous data messages.
Or the end-user must make many simultaneous associations with other end-users on the network.
For systems that have a low error rate, or for transmission in a reliable network with a low error rate.
Or a system where the order of the data blocks is automatically maintained.
It is not efficient.   To use the metropolitan area network efficiently, multiple user groups
It is hoped that the equivalent of private network service can be provided to each group
Good. Such a configuration is called a virtual network.   One problem in the prior art is that users of one virtual network can
Ensure that they do not gain access to private data accessed
Is difficult. An entity, such as a company, university or government agency
Even within, some data, such as salary records It is often required to restrict access to code and the like. Prior art systems
The system provides unauthorized access to virtual networks that are served by a common data network,
These networks have a common media architecture, and they have sufficient
It does not have sufficient ability to prevent authentication because it does not authenticate.   The solution to the above problems and the technical improvement from the prior art
Sends data packets containing user and port identification, and these packets are
Checked to see if the user / port pair is allowed
Is solved according to the principle of the present invention. As an advantage, this configuration is not
Users can use unauthorized access to identify unauthorized users from another port.
Prevent getting. According to one aspect of the invention, this port identification is added by the network.
And therefore are not under the control of the user. As an advantage, this is
Prevents bogus sources of data being transmitted over the network.   According to one aspect of the invention, the network places the destination port in a reserved location in the packet header.
Supply the default number. The advantage is that this destination port number is at the destination edge of the network,
It can be used to route this packet only to its destination port.   According to the present invention, it is checked that only users authorized to the network are communicating.
Responsibility for doing things. The header for each data packet is
Source and destination. If the source / destination pair is not allowed, the network will
Block the ket. The advantage is that only authorized source / destination pairs can communicate.
Wear.   The heading also identifies the physical port from which the data message is being transmitted.
Contains information to specify. This identification is provided within the network and allows the user to
Transmitted from one of the free ports in the restricted port group used in
It is difficult to falsely identify the object. The advantage is that only logged-in users
, Can transmit data packets with his user identification. This is the other All data blocks received by the network from the gateway will identify the user.
If not, it will be blocked in this network.   Headings are checked separately. If there is an error in the data in the heading
, The packet is discarded. As an advantage, individual checks of headings add additional security
I will provide a.   This protocol requires only a small amount of processing to detect errors.
Used with optical fiber short-haul networks with an error rate.
Is found, the data is controlled under the control of the end user terminal in response.
It is designed so that retransmission of the message is attempted.   In accordance with the principles of the present invention, network functions are more specifically performed in high speed
Data is collected at the edge of the network, and exchanged and
Is defined as a network function that provides a low error rate. The design of this net, in particular,
The data blocks in the order of the data messages are
Location Testify. This is how packets are processed in the nodes of the network and all connections.
Due to the fact that the continuation is exchanged only at the hub. In this network,
It is very rare that non-compliant packets are received, so the message
All you need is a simple user check to make sure it was successfully received.   In this embodiment, each user has an input with significant data buffering capacity.
Tied to the face. When data messages are received, they are
Buffered in the interface and allows the user to process the received data message correctly.
You are given time to allocate messages for processing. Heading is message length
Including data. As an advantage, this configuration allows data sending users to send data messages.
Receiver completes resource allocation for receiving the message
It is possible to send a notification of having done without waiting.   According to the present invention, the source, destination, and user groups of individual data entities are
Within the network, You are checked for proper permission. In one embodiment, the net further comprises
Authentication check by adding port identification at the beginning of each data entity, eg packet
Is also the entity coming from a legitimate port, and that port is the user
Can check if it is the same port that you logged into the system.
As an advantage, this configuration allows you to communicate with that destination group.
Login and password are granted, the
Only users transmitting from the specified port can communicate with the destination group
.   In one embodiment, the user's port is authorized for that user
Checked against the list of ports. The advantage is that this configuration can
Unauthorized users who gained passwords fraudulently by accessing the system from
Prevent users from acting as if they were legitimate users.   In one embodiment of the present invention, the user Upon logging in to the carrier network, the user has to identify himself, his password and his
Provides identification of specific virtual networks that the user wishes to access during the session
I do. The network verifies that this user has the right to access the virtual network.
For the user, this permission includes read-only access, read and write access,
Alternatively, it may be write-only access, but this carrier network
Identify unauthorized access requests based on security checks. As a result,
Only data packets for communication are transmitted by the network to their destination
. The advantage is that once a user's rights have been established,
All subsequent checks for messages from this user will be
Overhead, and there is almost no overhead at the end user.
Does not exist.   According to one aspect of the invention, a common group can be defined. This common group
Users who have access to the loop will be able to Members can also be accessed. This common group is sensitive
Additional password procedures in connection with the terminal, including users with information
Is required to access this information using common group facilities. Merit
As a result, you can use this configuration to
A simple connection to this is possible. Need this special property of private virtual network
Users who do not need to use only the services provided by the common user network
No.General description   The detailed description of this specification and Metropolitan et.
It is a description of a rear network (MAN). This net, as shown in FIGS. 2 and 3,
, A network interface module connected to the hub 1 by an optical fiber link 3
(NIM) 2 outer link. This hub can access data and data from any NIM
And voice packets to other NIMs. NIM, on the other hand,
Units connected to this network via Joules Connected to the user device.   The patent claimed here consists of a protocol used within the MAN
But this is added as a data packet entering the MAN and within the MAN
User can access the MAN only from the network port designated for that user.
And include parts of the protocol that are checked to ensure. Part of this claim
Parts of the detailed description that are closely related to Selections 9 and 10 and the first
5 and 20.   The invention claimed herein is compatible with networks such as MAN as described in the detailed description.
It also relates to the protocol used. Select this protocol for details
9 and 20. The transmission path of the network is described throughout the rest of the detailed description.
Explained. The data is transmitted to the user system and the data exchange module of the transmission side hub 1.
Module (MINT11) and the network interface module from the receiving user.
Depending on the specific choice of where in the UIM the data is stored,
Exchange and one-stage Having only circuit switching, as a result, the number of errors in transmission is minimized,
Packet order is always preserved, and therefore the protocol of the type described
A configuration is provided that allows use.   The type of MAN described in the detailed description serves multiple customers
With the ability to provide To make full use of these features, this network
For example, designed to handle multiple virtual networks dedicated to different businesses
Likely to be. In this situation, privacy between individual virtual networks
Users are not members of a particular virtual network.
  It is required to prevent access to the file. According to the principle of the present invention
This is the source port for each packet exchanged through the MAN network.
Identified so that only those ports associated with one virtual network can be
Access to the site.   The patent claimed herein provides a mechanism for providing virtual network services to a group of users.
About. This machine The scheme uses the authorization data 360 and the MINT central control 20 (14th
Source checker table 308 and routing table 31 in FIG.
This is realized by inputting data into 0. Selection 10 is especially virtual
In describing the implementation of the network, selection 9 describes the protocol.
You.Detailed description 1. Introduction   Data networks are usually categorized by their size and range of owners. locker
Le Area Networks (LANs) are usually owned by a single organization,
It has a spread. These are dozens or hundreds of terminals, computers, and other end users.
Interconnect the system (EUS). At the other extreme, there is a spread between continents
There are two wide area networks (WANs), which are
Owned by talking companies, interconnect tens of thousands of end-user systems (EUS)
. Between these extremes, the range extends from campus Other data networks leading to the metropolitan area are identified. Explained here
The high performance metropolitan area network is called MAN. Initials and abbreviations in Appendix A
The table is given.   The metropolitan area network is a simple reporting device and low intelligence
Computers, and large mainframes and supercomputers
Provide services to various EUS. These EUS ask the net
Services are very mixed. Some EUS issues messages very rarely, some
EUS issues a number of messages every second. Some messages are several bunts long
Only. Some messages consist of multiple files of millions of bytes. An E
The US requires delivery at any time within hours, and some EUS require microseconds
Demands delivery within.   Metropolitan area network according to the present invention realizes broadband low-latency data transmission
Computers and telephone networks designed to perform
Local area network performance characteristics Or more. The metropolitan area network is class 5, or end-office.
Has size characteristics comparable to end-office telephone central offices;
Now consider the metropolitan area network as the end office for data
Can be. An exemplary embodiment of the present invention, hereinafter referred to as MAN, makes this fact possible.
Designed with your mind in mind. However, MAN is a replacement module for the end office.
Broadband integrated services designed as an accessory or part of
Supports digital network (Integrated Service Digital Netwprk, ISDN) services
You can also. MAN can also be in the local or campus area
It is also effective as a net. This can be done from a small LAN through a campus-sized network.
And grow gracefully into a full MAN.   The rapid growth of workstations and these servers, and the growth of distributed computing
This was a major motivation for the design of the present invention. MAN is tens of thousands of Dixless Works
Stations and servers and other Designed to connect computers through dozens of kilometers. Individual
The user has thousands of simultaneous different associations with other computers on this network. Individual
Each computer connected to the network can send tens to hundreds of messages per second at the same time.
Simultaneous generation and I / O speeds of tens to hundreds of million bits per second (Mbps)
Request. Message sizes range from hundreds of bits to hundreds of bits. this
MAN requires remote procedure calls, object
Communication, remote request paging, remote swapping, file transfer, and
Has the ability to support computer graphics. The goal is most messages
(Hereinafter referred to as a transaction) from one EUS memory to another EUS memory
Within 1ms for small transactions and large transactions
Then, the transmission is within a few milliseconds. Figure 1 shows the transaction type
Categorize the required EUS response time for both transaction type and size.
As a function of You. Simple (ie, low intelligence) terminal 70, remote procedure call (RPC) and object
Inter-project communication (IOC) 72, paging on demand 74, memory swapping
76, video computer graphics 78, still computer graphics
, Computer for file transfer 82, and packetized audio 84
Network requirements are indicated. The response time / transaction shown in FIG.
Satisfying speed is one of the goals. As a scale, a constant bit rate
This indicates that this bit rate dominates its response time.
Show. The MAN has a total bit rate of 150 Gbit / s and is shown in FIG.
If an example processor element is selected, a network of 20 million per second
Can handle transactions. In addition, this favors traffic overload.
It is designed to be able to handle elegantly.   A MAN is a network that performs switching and routing like many systems,
This is even more Various other necessary functions such as error handling, user interface, etc.
Have. Authentication capabilities provide excellent privacy and security within the MAN
Have been. This feature prevents unauthorized use of the network and reduces usage-sensitive charges.
Money billing (usage sensitive billing)
Non-forgeable source identification is provided. In addition, a virtual private network has been established.
It has a function to define.   MAN is a transaction-oriented (ie, connectionless) network.
You. This means that you can add a connection veneer if necessary.
But does not incur the overhead of establishing or maintaining a connection.   MANs can also be used to exchange packetized voice. Network transit delay
Is short, transmission priority is given to a single packet, and there is no heavy load on the network
Due to the small fluctuation of delay at the time, voice or a mixture of voice and data is sent to MAN.
It can be easily supported. For simplicity, As used herein, the term data refers to digital data representing an audio signal,
Displays instructions, numerical data, graphics, programs, data, and files.
Digital data, and other contents of the memory.   MAN is not fully built, but extensive simulations have been performed.
It has been done. Many of the performance estimates presented here are based on these simulations.
Good.2. Architecture and operation 2.1 Architecture   The MAN network can be two or three levels depending on how close it is viewed
It is a hierarchical star architecture with. Figure 2 shows this network at the edge of the network.
Called hub 1 linked to network interface module 2 (NIM)
Shown as consisting of an exchange center.   This hub is a very high performance transaction-and-forward system.
Stems, which can be very large Can handle more than 20 million network transactions per second, 1 per second
Gracefully grow to a system with a total bit rate of 50 Gigabit.   From this hub, it is called an external link (external link, XL).
Fiber (up to tens of kilometers) (or connecting to INT)
Alternate data channels) radiating, each at full duplex bit rate per second
It has the ability to handle on the order of 150 megabits. XL terminates at NIM.   NIM, whose outer edge outlines the edge of the network, is a concentrator / demultiplexer
It works with both devices and identifies network ports. This means that when information is transmitted through the network
When the information is transmitted outside the network, the signal is separated. Collection /
The purpose of the separation is to link multiple end user systems 26 (EUS) to the network efficiently.
Interface in such a way as to be used economically and economically. EU
Depends on network demand of S but individual NIM Up to 20 EUS26s. Examples of these EUS
There are workstations 4 with advanced functions that are becoming increasingly common
However, this burst rate is already in the range of 10Mbps (this allows faster system
Is a matter of time), and has an average velocity one order of magnitude lower. EUS is
Requires an average speed close to that of NIM, and the average speed is the same as that of NIM.
In the case of the size of a maker, NIM has multiple interfaces to one EUS26.
To provide one interface or all of NIM and XL.
The body can be dedicated to the EUS. As an example of this type of EUS
Is a large mainframe 5 and file
6 and local area networks such as ETHERNET8 and Proteon
p.) of Proteon 80, an 80Mbit token ring manufactured by
A high-performance local area network 7 or a developing American Standards Association
(ANSI) Fiber Distributed Data Interface, the standard protocol ring interface for
Interface (FDDI). In the latter two cases
In other words, the LAN itself collects data, and the NIM is a single port network interface module.
Can also be degenerated by a tool. Lower performance local area network, eg
For example, ETHERNET8 and IBM Token Ring are all of the capabilities offered by NIM as a whole.
Not needed. In such a case, this LAN will collect data,
It can also be connected to port 8 on the NIM.   A user interface module (UIM) 13 exists in each EUS.
Exist. This unit provides a high bit rate direct memory access point to the EUS.
Function as a buffer for transactions received from the network and the network.
It also undertakes MAN interface protocol issues from EUS. M
An EUS-resident driver exists in close relationship with the UIM. This is a collaborative transaction with UIM for outgoing transactions. Formatting, receiving incoming transactions, realizing protocols, and EUS
Interfaces with the operating system.   A closer look at the hub (see Fig. 3) reveals two different functional units,
MAN switch (MANS) 10 and one or more memory interfaces
It can be seen that the base module 11 (MINT) exists. Individual MINT
Are connected via XL3 to up to four NIMs, thus up to 80 EUs
Selection of 4 NIMs per MINT capable of accommodating S transaction processing capacity
, Buffer memory size in MINT, network growth capacity, failure group size
Size (failure group size) and total bit rate.   Each MINT has four internal links 12 (IL) (MINT and MAN switches).
Connected to the MANS, one of which is an individual of the MINT of FIG.
Shown against. The reason four links are used in this case is XL
It is different from the case. Here, MINT normally It sends to multiple destinations simultaneously through the MANS, and in a single IL this is the bottleneck
Therefore, multiple links are required. Selection of four ILs (as well as similar
Many other design choices of nature) are available for a wide range of analysis and simulation models.
Is based on the IL runs at the same bit rate as the external link,
Very short because the entire hub is co-located.   The smallest hub consists of one MINT, IL loops back,
There is no switch. A network based on this hub contains up to four NIMs and
Accommodates 80 EUS. The largest possible hub at this time is 256
It is composed of MINT and 1024 × 1024 MANS. This hub is 10
Accommodates 24 NIMs and up to 20,000 EUSs. Add MINT
Well, by growing MSNS, this hub and eventually the entire network
Show elegant growth.2.1.1 LUWU, Packet, SUWU, and TOW Transaction   Before proceeding, some terms need to be explained. EUS transactions
This is the transmission of EUS information of a unit that has meaning for EUS. This transac
The option can be a remote procedure call consisting of several bytes or a 10 megabyte
It could also be a database transmission. MAN has, for the purposes of this description,
User work unit (long user work uint, LUWU) and short
  2 called short user work unit (SUWU)
Identify two EUS transaction unit sizes. Determination of size range
Although this is technically simple, transaction units typically have thousands of bits or less.
Is considered a SUWU and larger transaction units are LU
Considered WU. The packet is based on the criteria shown in Figure 1 in the network.
Priorities are given to reduce response time. Fig. 1 shows a small EUS tiger
Transaction units are usually faster It can be seen that an EUS transaction response time is required. These packets are
Is transmitted as a single frame or packet when transmitted
Keep in shape. The LUWU transmits a frame called a packet hereinafter by the transmitting UIM.
Or divided into packets. Packets and SUWU are often collectively network traffic
Called a transaction unit.   Transmission through the MAN switch is called a switch transaction,
Units transmitted through the NS are called Switch Transaction Units.
Devour. These are one or more network transactions directed to the same MIN.
Configuration unit.2.2 Overview of functional units   Before describing the operation of the MAN, an overview of each of the major functional units in the network is given.
Need to be The units described here are UIM13, NIM2, MINT
11, Mans 10, End user system link (connect NIM and UIM) (EUSL) 14, XL3, and IL12. These units are shown in FIG.
Is done.2.2.1 User Interface Module-UIM13   This module is located in the EUS and is usually Plug into an EUS backplane such as a channel. Most applications
Is designed to be mounted on one printed circuit board. UIM
13 is provided by optical transmitters 97 and 85 called EUS link 14 (EUSL).
Connected to MIN2 through a dual optical fiber link driven by This phosphorus
The link runs at the same speed as the external link (XL) 3. UIM is on the way to the network
And has a memory queue 15 used to store information here. Packet
And SUWU use out-of-band flow control.
Transferred to the NIM.   Receive buffer to receive information from the network A memory 90 is required. In this case, the entire EUS transaction
Are often stored before they are transmitted to end-user system memory
. The receiving buffer is required to have dynamic buffer connection capability. Partial EU
S transactions may arrive in a concurrently inserted form.   The optical receiver 87 receives a signal from the optical link 14 and converts it into a reception buffer memory.
It is stored in the memory 90. Control 25 controls UIM 13 and
First queue (FIFO) queue 15 or receive buffer memory 90 and end user
A bus interface for interfacing with a bus 92 connecting to the system 26;
Controls the exchange of data with the source. Details of control of UIM13 are shown in FIG.
It is.2.2.2 Network Interface Module-NIM2   NIM2 is part of the MAN and resides at the edge of the network. NIM is
Six functions, (1) Queuing of packets and SUWU moving toward MINT and
Concentration / signal separation including arbitration of external links, (2) Network security using port identification
(3) Participation in traffic congestion control, (4) EUS to network control
Message identification, (5) participation in error handling, and (6) network
Perform interface operations. Similar to outgoing FIFO 15 found in UIM
A small queue 94 in memory exists for each end-user system.
. They receive information from the UIM via link 14 and receiver 88,
Store until XL3 becomes available for transmission to MINT. These queues
Drive the data concentrator 95, while the concentrator 95 drives the optical transmitter 96.
You. There is an external link request multiplexer, which is required for the use of XL.
Answer the request. NIM sends port identification number 600 (Fig. 20) to MINT.
At the beginning of each network transaction unit that is This is different
By law, added value Services, for example, ensure reliable and fake sender identification and billing behavior
Used to This prefix is used to identify members in the virtual network
Necessary to protect against unauthorized access. Inspection sequence error control
Processed for the rules. MIM detects congestion in the network in cooperation with hub 1
Then, when traffic congestion becomes significant, the flow from the UIM is controlled. NIM is also a network
Provides standard physical and logical interfaces, including flow control mechanisms.
Offer.   The information flowing from the network to the EUS is passed through the NIM via the receiver 89,
Distributor 86 is distributed to the correct UIM and transmitted to the destination UIM 13 by the transmitter 85.
Via link 14. There is no buffering at the NIM.   There are only two types of NIMs. One type (in FIG. 4 and in FIG. 3)
The type shown at the top right performs concentrating and the other type (shown at the bottom right of FIG. 3)
Does not do this.2.2.3 Memory and interface modules -MINT11   MINT is located in the hub. Each MINT11 terminates (a) XL,
Also, a signal is sent from half of the internal link that moves data from switch 10 to MINT
Up to four external link handlers 16 (XLH) that receive the
4 data generating data for half moving data from MINT to switch
Internal link handler 17 (ILH); (c) N from MINT through switch
Memory 18 for storing data while waiting for a route to the IM; (d) link
  Data that moves data between the handler and memory and also carries MINT control information
And (e) a control unit 20.   All functional units in the MINT may enter or be in the MINT at the same time.
It is designed to accommodate the peak total bit rate for outgoing data. This
This ring, which is synchronous, carries information from individual XLHs to memory
Set of reserved slots, and information from memory It has another set of reserved slots to carry to individual ILHs. this
Has a read plus write bit rate of 1.5 Gbps or more. This memory is 51
2 bits wide and therefore sufficient memory with elements with reasonable access times
Bit rates can be achieved. The size of this memory (16 Mbytes)
Because the information is placed for a short time (approximately 0.57 milliseconds at full network load),
This size is changeable and can be adjusted if necessary.
I can do it.   XLH is bidirectional, but not symmetric. Information to move from NIM to MINT
The information is stored in the MINT memory. The heading information is copied by XLH,
Sent to MINT control for processing. Conversely, switch 10 to NI
Information moving toward M is not stored in MINT on the way, but simply processed.
Without passing through MINT from the output of MANS10 to the destination NIM2.
Due to the variable path length in the switch, the information leaving the MANS 10 is phase-sensitive to XL.
Gaze It is. N in the phase matching and scrambler circuit (described in section 6.1)
The data is aligned before transmission to the IM. About Internal Link Handler (ILH)
Is described in section 4.6.   MINT has (1) some overall routing in the network, (2) user verification
, (3) participation in network security, (4) queue management, (5) network transactions
(6) Address translation, (7) Participation in traffic congestion control, and (
8) Various functions including generation of operation, management and maintenance (OA & M) primitives
Perform   Control for MINT is in accordance with MINT control algorithm
It is a data flow processing system designed by Individual MINTs are
It has the ability to handle up to 80,000 network transactions. 250 MINTs
A fully-equipped hub that has, therefore, 20 million network transactions per second
Can be processed. See section 2.3 for this. Further explained.2.2.4 MAN switch-MANS10   The MANS is composed of (a) an organization 21 through which information is passed, and (b)
It consists of two main parts consisting of controls 22 for the weave. This con
Troll allows the switch to be set up within about 50 microseconds
You. Due to the special characteristics of this organization, the control can be operated in parallel
Can be disassembled into independent Zab controllers. In addition to this, individual
Sub-controllers can be connected to the pipeline. Thus, set
Not only is the up-time very fast, but also multiple
The “setup throughput” can meet the high demand speed from many MINTs.
Raised high enough to accommodate. MAN is 16x16 (processes 4 MINTs)
To 1024 x 1024 (processing 256 MINTs)
It can be designed in size.2.2.5 End user system link- EUSL14   The end user system link 14 locates the NIM 2 in the end user's device.
Connected to the UIM 13. This is a full-duplex fiber optic link, which is
Runs synchronously at the same speed as the external link on the other side. This is the E to which it is connected
Used exclusively for US. The length of EUUSL is several meters to several tens of meters.
Is assumed. However, if it is economically feasible, no longer
There is no reason.   Basic format and data for EUSL for this embodiment of the invention
The speed was measured using the Metrobus Lightwave System OS-1 Link.
OS-1 link) was selected identically to this. Eventually any link layer data transmission
Even if a standard is adopted, it can be used in later embodiments of the MAN.2.2.6 External link-XL3   An external link (XL) 3 connects the NIM to the MINT. This is also a full-duplex synchronous optical
With fiber link is there. This is required by the end-user system connected to the NIM.
(demand multiplexed fashion). XL is several tens of kilometers long
Is assumed. Request multiplexing is used for economic reasons. this
Employs the Metrobus OS-1 format and data rate.2.2.7 Internal link-IL24   An internal link 24 provides a connection between the MINT and the MAN switch. this is
A unidirectional semi-synchronous link that retains frequency when passing through MANS10
However, the synchronous phase relationship is lost. Length of IL24 is assumed to be on the order of several meters
However, it can be longer if economically feasible. IL bit
The bit rate is the same as the bit rate of OS-1. However, the format is data
Has only limited similarity to OS-1 due to the need to resynchronize.2.3 Software Overview   Workstation / server paradigm Each end-user system used and connected to the MAN has LUWUs per second
Can generate more than 50 EUS transactions consisting of SUM and SUWU
Noh. This is 400 network transactions per second (packets and SUW
U). Having up to 20 EUSs per NIM is a matter of individual NI
M handles up to 8000 network transactions per second, and each MINT
4 times the number of network transactions, or 32,000 network transactions per second.
Means you have to. These are average or sustained rates. bar
Depending on the strike condition, the "instantaneous" speed for a single EUS 26 can be greatly increased.
There is. However, individual EUS bursts are averaged over multiple EUSs.
Can be smooth. Therefore, individual NIM ports should be more than 50 per second
NIM (NIM)
2) and XL (3) are not expected to encounter too many bursts
. This is an individual Is even more pronounced for MINTs that process four NIMs. MAN switch 10
Must pass an average of 8 million network transactions per second
However, the switch controller is not required to handle this many switch requests.
Not required. This is due to the design of the MINT control towards the same destination NIM.
Multiple packets and SUWUs are exchanged in a single switch setup
It is because it is devised as follows.   The second factor to be considered is the network transaction interarrival time.
me). 150Mbps speed and the smallest network of 1000 bit SUWU
Assuming a transaction, the two SUWUs become NIM or MINT
6.67 microseconds apart. NIM and MINT are transient
Must be able to handle several folded SUWUs.   The control software in NIM and especially in MINT,
Real-time transaction processing. Day Peak load is processed in a single period due to asymmetric and bursty characteristics of traffic
A design that can be done is required. Therefore, transaction control software
The hardware structure can execute hundreds of millions of CPU instructions per second.
Ability is required. In addition, within the MAN, this control software
The routing of packets and SUWUs, identification of network ports, up to 1000
Queuing of network transactions directed to up to the same NIM (this is
Real-time holding of up to 1000 queues), MANS requests and receipts
Knowledge processing, outbound EUS flow control based on complex criteria, network traffic
Various tasks including data collection, traffic control, and numerous other tasks.
Perform a function.   MAN Control Software performs all of the above tasks in real time
Have the ability to perform. This control software is NIM Control
23, MINT control 20, And the MANS control 22. these
In conjunction with the three control elements, in the UIM 13 of the end user system 26
Has a fourth control structure 25. FIG. 5 shows this configuration. Individual N
IM and MINT have their own control units. These controls
  Units work independently but work closely together. Of this control
Partitioning is an architectural enablement of MAN's real time transaction processing capabilities
This is one of the texture mechanisms. Allow MAN to handle high transaction rates
Another mechanism that works is to break this control down into a logical array of sub-functions.
This is a technique for independently adding processing power to individual sub-functions. This approach is
INMOSGreatly assisted by the use of the device. This technology is basically as follows
It is.   -Break down the problem into multiple sub-functions.   -Combine these sub-functions into one data flow structure It is arranged to form.   -Implement individual sub-functions as one or more processes.   -Binding a set of processes to a processor, and binding the bound processor to its data
Data flow that performs this function, arranged in the same topology as the data flow structure
  Forming system.   − Iterating as needed to achieve the required real-time performance
.   The functions performed by NIM, MINT, and MANS (most of these are
(Performed by software for these modules).
Given in 2.2 to 2.2.4. Additional information in section 2.4
Has been given. These in specific sections covering these subsystems
Will be described in detail.2.3.1 Control Processor   The processor selected for the implementation of the system was Transp from INMOS.
uter. this 10 million instructions / second (MIP) shortened instruction set
The reduced instruction set control (RISC) machine
Designed to connect to any topology via 0Mbps serial link
Have been. Each machine has input and output channels with simultaneous direct access (DMA) capability.
Includes four links with paths.2.3.2 Performance of MINT control   Due to the need to process a large number of transactions per second,
The processing of an action is broken down into serial sections that form a pipeline.
Transactions feed into this pipeline, where they are
It is processed at the same stage as other transactions at the same stage. In addition to this,
There are multiple parallel pipelines, each handling its own processing stream at the same time.
Exist. In this way, individual transactions can be routed and other complex services
The desired high transaction processing speed that requires Such a parallel / pipe of a processor with interconnected control structures
Achieved by breaking down into in-connected structures.   Constraint on MINT control, arbitrary serial processing is given by the following formula
Must not be longer than   1 / (number of transactions processed per second in this pipeline)   Another limitation is that headings that enter controls in XLH16
There is bandwidth. The spacing between subsequent network units to reach XLH is:   (Heading size) / (Bandwidth to control) If shorter, the XLH is required to buffer headings. Imagine a uniform arrival
The maximum number of transactions per second at a given time is given by:   (Bandwidth to control) / (Transaction heading size)   Effective bit rate of transputer link and 40 byte MAN
An example based on the network transaction heading is as follows.   (8.0 Mb / s for control link) / (320 bit header / to
Transaction) = 25.000 transactions / sec / XLH,   That is, 1 transaction / XLH per 40 microseconds. Transa
Transaction interarrival times can be shorter
Because of this, heading buffering is performed within the XLH.   MINT will execute the routing and billing primitives during this time.
Line, switch request, network control, memory management, operation, supervision, and
Perform maintenance activities, provide name services, and add Yellow Pages
Other network services, such as mobiles, must also be provided. MINT Con
These goals are achieved by the parallel / pipeline connection characteristics of Troll 20.
You.   As an example, allocation and deallocation of fast memory blocks can be routed or
Is processed completely independently of the billing primitive. Transa in MINT
An action flow is a network transaction unit (that is, a packet or
SUWU) Memory block address management program used to store
Controlled in a single pipe by the ram. In the first stage of this pipe,
The memory management program allocates empty blocks in the high-speed MINT memory. Next
In the next stage, these blocks are paired with headings and
Is performed. Next, the switch unit sends the memory block sent to the common NIM.
Is collected based on the lock, and this block of data is transmitted to the MANS.
After the memory block is released, the loop is closed. Billing pre
Mititives are processed simultaneously in different pipes.2.4 MAN operation   EUS26 is controlled by a network management program from the network. It is considered that the user has the ability given. This idea is logged in a time-sharing system.
Similar to the end user that was plugged in. Users, eg stations or even
Is a workstation or preprocessor that acts as a concentrator for the network.
The server makes a physical connection at the NIM port and then its MAN name, virtual network
Identify and identify yourself through password security.
Required. The network stores the routing table with data destined for this name.
Adjust to map to the desired NIM port. These features for this user
Associated with a physical port. The example shown here is a portable workstation
Houses the paradigm of Yeon. The ports are also configured to have fixed functions and
In some cases, it may be "owned" by the end user of one MAN designation
You. This gives the user a dedicated network port, or a privileged maintenance port (privileged
Prepare ged administrative maintenance port). Outgoing EUS is destination MAN
Name or service And therefore they know nothing about the dynamic network topology.
Not required.   Are the high bit rates and large transaction throughput in this network very short?
Response time and transfer EUS data without excessive network considerations.
To provide a means for As a result, the response from the EUS memory to the EUS memory
Time is as short as 1 millisecond, low error rate, and sustained base
Can carry 100 EUS transactions per second. This number is
Scalable to thousands of high performance EUSs. EUS is an unlimited user
You can send data in any size you want. In optimizing MAN performance
Most of the restrictions are not on network overhead, but on EUS and application
Defined by constraints. The user provides the following information when transmitting data to the UIM.
Supply.   The MAN name and the virtual network name for the destination address independent of the physical address.   The size of the data.   A MAN type field indicating the required network service.   -Data.   Network transactions (packets and SUWUs) move along the following logical paths
(See FIG. 5).   Outgoing UIM → Outgoing NIM → MINT → MANS → Destination NIM (via MINT
T) → destination UIM   Each EUS transaction (that is, LUWU or SUWU)
Sent to UIM. Within the UIM, the LUWU also has a variable-size packet.
Fragmented. SUWU is not fragmented and logically consists of one network.
Considered a transaction. However, one network transaction is one SU
The determination as a WU is not made until the SUWU reaches the MINT. MI
In NT, this information is used to make the data SUW for optimal network bundling.
It is dynamically classified into U and packet. NIM uses incoming packets from EUS
To see if it violates the maximum packet size. UIM is a service specified by EUS.
Than the maximum size determined by the screw You can also decide on a smaller packet size. Use of optimal MINT memory
For the packet size, the standard maximum size is appropriate. However, depending on the situation,
End user considerations such as timing issues or data availability timing.
May require the use of smaller packet sizes. In addition to this,
UIM is timing related to the situation of transmitting what is currently received from EUS
There are also restrictions. Even if the maximum size packet is used, the last packet of the LUWU
The packet is usually smaller than the maximum size packet.   Individual network transaction (packet or SUWU) at the sending UIM
Is prefixed with a fixed-length MAN network header. MAN Network Software Found
Using this information in the routing, billing, provision of network services, and network
Control. The destination UIM uses the information in this heading to
Perform the job of delivering the transaction to the end user. Network transaction
Is in the UIM outgoing transaction queue , Which are sent from here to the originating NIM.   When receiving network transactions from the UIM, the NIM places them there.
Stored in a queue permanently dedicated to the EUSL that the transaction arrives at,
Later, this is transmitted to the MINT 11 as soon as the link 3 becomes available. In NIM
Control software to handle UIM → NIM protocol
Therefore, the control message is identified and the outgoing port number is transacted.
To the head. This is used by MINT to authenticate this transaction
. End-user data is the content that the data is provided to the network by the end-user.
Unless addressed as troll information, some MAN network software
Never touched. Once these transactions have been processed, the originating NIM
Collect them on the external link between the originating NIM and its MINT. Outgoing NIM
The link from MINT to MINT is provided by a hardware interface (external
Nku Handler, ie, XLH16).   The external link protocol between NIM and MINT is based on XLH16 network
It is possible to detect the start and end of a transaction. These transactions
Now handles 150 Mb / s bursts of data arriving at XLH immediately
It is moved into the designed memory 18. This memory access is fast time,
It is performed via a lot ring 19, wherein the ring 19 has an individual 150 Mb / s
XLH input and individual 150 Mb / s output from MINT (ie, MANS input)
Force) guarantees bandwidth without contention. For example, collecting four remote NIMs
, MINT with 4 input ports to central switch is at least 1.2 Gb / s
Must have a burst access bandwidth of Memory storage
Fixed size equal to large packet size plus fixed length MAN header
Used for long blocks. XLH is the address of a fixed-size memory block
And subsequent packets or SUWU data Data to this memory access ring. Data and heading are MIN
It is stored until transmission to the MANS is instructed by the T control 20. MI
NT control 20 stores this incoming packet and SUWU in XLH
Provide uninterrupted empty memory block addresses. XLH is also solid
"Know" the length of the fixed-size headline. On the basis of this information, XLH
The copy is passed to the MINT control 20. MINT Control 20
Is a block that has been given to the XLH to store this header and the packet or SUWU.
Paired with the IP address. This heading is the data in the MINT control
Must be absolutely correct because it is the only internal representative of Link Ella
In order to ensure hygiene from the public, etc., headings should be checked for their own cyclic redundancy check (CRC
)have. The route that this tuple passes through the MINT control is arbitrary
Must be the same for all packets of the LUWU
USU data Order is preserved). The MINT memory block address and the paired
The packet and SUWU move through the processor pipeline. This pipeline is
Network transactions where multiple CPUs are different at different stages of MINT processing
Can be processed. In addition, there are multiple pipelines,
Thereby, simultaneous processing is realized.   The MINT control 20 uses the unused internal link 24 from this IL.
Request to set route to destination NIM (through MINT connected to this NIM)
I do. The MAN switch control 21 places this request in a queue and (1)
When the route of (2) becomes available, and (2) XL3 to the destination NIM becomes available,
Notify INT and set this route at the same time. This is flat, under full load
Requires 50 microseconds. Upon receipt of the notification, the originating MINT sends its N
Sends all network transactions destined for the IM, and thus
Path is optimized You. The internal link handler 17 requires a network transaction from the MINT memory.
And transmit them through the following path.   ILH → outgoing IL → MASN → destination IL → XLH   This XLH is connected to the destination NIM. XLH bit on the way to destination NIM
Restore synchronization. The information is that when it leaves the switch, the MI on the way to the destination NIM
Note that it simply passes NT. MINT passes through MANS
No information processing is performed except for recovering the bit synchronization lost in the process.   Information (ie, a switch consisting of one or more network transactions)
When the transaction arrives at the destination NIM, it
And SUWU) and transmitted to the destination UIM. This is "on the
In other words, it is stored in the NIM on the way before leaving this network.
It will not be.   The receiving UIM 13 stores this network transaction in its receiving buffer memory 90.
And EUS Replay transactions (LUWU and SUWU). LUWU sends a message to UIM.
Reached in ket size fragments. Just when at least part of the LUWU arrives
The UIM informs the EUS of this existence and, upon receiving a command from the EUS,
Under the control of DMA, part or all of the EUS transaction
It is sent to the memory at the DMA transmission size specified by EUS. From UIM
An alternative paradigm for transmission to EUS can also be used. example
If the EUS reaches the UIM in advance, it immediately writes this to the EUS memo.
You can also tell it to transmit in a buffer specified in the directory. in this case,
The UIM does not need to notify the arrival of the information, but transmits it to the EUS immediately.
Can be.2.5 Additional considerations 2.5.1 Error handling   Achieved latency of several hundred microseconds from EUS memory to EUS memory
To do so, the errors are different from those used by traditional data networks. It is necessary to handle it in some way. In MAN, network transactions
The heading check sequence 626 (FIG. 20) (HCS) and the
Data inspection sequence 646 (FIG. 20) attached to the entire network transaction.
) (DCS).   First consider the headings. The originating UIM must be HS
Generate C. At MINT, HCS was checked and an error was found
This transaction is discarded. Destination NIM performs similar actions three times
After that, the transaction is sent to the destination UIM. This scheme looks down
Prevent erroneous delivery of information due to delivery. If any defects are found in the headline,
Are considered unreliable, the MAN discards this transaction
Have no other choice.   The originating UIM or at the end of the user data shall provide a DCS.
Required. This field is checked in the MAN network. However, No action is taken if an error is found. This information is delivered to the destination UIM
Are checked here and appropriate actions are taken. This use in the network is EUS
The problem is to identify both L and internal network problems.   Within this network, the normal autonomy found in most of today's protocols
No attempt has been made to correct errors using the Automatic Repeat Request (ARQ) technique.
And be careful. This is not possible due to requirements from low waiting times. Error correction skis
Systems are too expensive, except for headings, and even in this case,
The time penalty is too great, as is often the case in Tem. However
Demonstrates that heading error correction is empirically necessary and possible in time
If so, it can be adopted later.   Therefore, if the MAN checks for errors and has reason to doubt the validity of the heading
Destroys the transaction. Otherwise, transactions are flawed.
Even if it is delivered. This is This approach is meaningful for three reasons. First, the normal ARQ protocol
Is adopted, the inherent error rate through the optical fiber is the error through the copper wire.
Same order as rate, both are 10^-11Bit / bit range. No.
Second, graphics applications (which are growing rapidly) are usually small
1 or 2 bits per image when the pixel image is transmitted
Does not matter. Finally, it is required that the error rate be better than the intrinsic rate.
In such cases, use the EUS-> EUS ARQ protocol to achieve this error rate
It is possible to2.5.2 Authentication   The MAN provides an authentication function. This feature allows the destination EUS to send
Guarantee the identity of the originating EUS for all transactions. Malicious user
Cannot send a transaction with a false "signature". The user is also a web
From being used for free. All users are sent to the individual
All transactions Is forced to identify itself in a true
Usage-sensitive billing is ensured. This feature also supports virtual private
It provides primitive capabilities for other functions such as networking.   When the EUS first connects to the MAN, this is a well-known privilege that is part of this network.
"Log in" to the login server. This login server is connected
Resides in a management terminal 350 (FIG. 15) having a disk memory 351. This
Management terminal 350 is the OA & M MINT professional in MINT central control 20.
Sessa (Figure 14) and MINT OA & M Monitor 317 and OA & M Central
Control (FIG. 15). This login is by EUS
Sends a login transaction to the server over the network (via its UIM)
Is achieved by This transaction includes the EUS identification number (the
Name), the required virtual network, and the password. Within NIM,
Port number is After being added to the head of this transaction, this
Transmitted to INT. The login server looks up this identification / port pair and
The MINT connected to the communication NIM is notified of this pair. This is also a login
Informs the EUS that it has received it, so that it may use the network.
I tell.   If this network is used, all individual network traffic sent from the EUS to the receiving NIM
The transaction shall include, in its headline, in addition to its source identification, a link to Figure 20.
And other information in the heading, which will be described in detail below. NIM tracks port numbers.
In addition to the head of the transaction, transmit this to MINT, where this pair is checked
Is done. If there is an invalid pair, MINT discards this transaction
I do. In MINT, the outgoing port number prefixed is the destination port number.
Replaced and then sent to the destination NIM. The destination NIM uses this destination port number.
To complete the routing to the destination EUS.   If the EUS wants to disconnect from the network, this is done in a manner similar to login.
The login server notifies MINT of this fact, and the MINT
Removes its identification / port information, which causes the port to be deactivated.2.5.3 Order guarantee   There is no LUWU concept from NIM to NIM. This NIM → NIM envelope
Within a LUWU loses their identity, but any LUWU's packet is
One path must be taken through L and MINT. This allows U
The order of the packets arriving at the IM is stored for a certain LUWU. However,
Some packets may be discarded due to defect headings. UIM is lost
Check the packet and notify EUS if this occurs.2.5.4 Virtual circuits and finite LUWU   This network does not set up a single circuit to the destination, but rather a group of packets and S
The UWU switches as resources become available. However, this is It does not prevent EUS from setting up virtual circuits. For example, EUS is suitable
Infinite size LUWU can be written with appropriate UIM timing parameters
it can. Such a data stream appears to the EUS as a virtual circuit, and for the network,
Looks like an endless LUWU carrying packets one at a time. Of this concept
The implementation is based on UIM and EUS because there are many different types of EUS and UIM.
Must be handled between protocols. End users may have multiple addresses
Multiple data streams may be sent first. These data streams are the originating UIM and the originating NI
M are multiplexed into packets and SUWUs at the boundary on the transmission link between M.   The parameters are adjusted for optimal performance when the system is loaded.
This limits the time one MINT can be sent to the NIM (data
To limit the length of the flow), which allows the NIM to be
Release for data reception. According to simulation, initial value of 2 ms
Look like is appropriate I can. This value is dynamically adjusted according to the traffic patterns in the system.
For different MINTs or NIMs, or at different times,
Alternatively, different values can be used for different days of the week.3. switch   MAN switch (MANS) is a high-speed circuit switch located at the center of the MAN hub
It is. This interconnects MINTs and all end-user transactions
You have to go through this. MANS is the switch organization itself (data network or
Is called a D-net) and a switch control complex (SCC).
In other words, it is composed of a set of controllers and links operating the D-net organization.
The SCC receives the request from the MINT and enters and exits the pair on the internal link (IL).
Connect or disconnect, make these requests when possible, and
Notify the MINT of the result of the request.   Perform these seemingly simple operations at a high performance level Is required. Requirements for MAN switches are described in the next section.
Explained. Next, in Section 3.2, the resolution of this switch requirement
The basics of a distributed control circuit switching network provided as the basis for C
In section 3.3, this approach applies to the specific requirements of the MAN.
And some features of this control structure that are important for high performance are described.
You.3.1 Characterization of the problem   First, some numbers for the demands on the MAN switch are estimated. nominal
Above, the MANS run individually at 150 Mb / s, each with thousands of separates per second.
Within a network with hundreds of ports carrying transactions that are switched
To establish or disconnect a transaction in a fraction of a millisecond
Is required. A transaction request of several millions per second can result in multiple
Transactions targeted by the pipelined control are
Larrel Means that a distributed control structure to be sequenced is required.   Many constraints to join so many ports, each running at high speed
There is. First, the network bandwidth must be at least 150 Gb / s.
Requires multiple data paths (150 Mb / s nominal) through this network.
Second, a 150 Mb / s synchronous network (asynchronous networks require clock and phase recovery)
However, construction is difficult. Third, in-band signaling is more complex (self
Routing) Out-of-office to provide network organization and require buffering in the network.
A band signaling (individual control) approach is desirable.   In MAN, transaction length is expected to fluctuate by several orders of magnitude.
It is. These transactions are, as explained later, small transaction
Share a single switch with sufficient delay performance for the switch. Benefits of a single organization
Separates the data stream before the exchange and does not need to rejoin it after the exchange
That is.   A problem to consider is when the required output port is busy. Contact
To set up a connection, any input and output ports must be idle at the same time.
No (it has a so-called concurrency problem). An idle input (output) port
Waiting for the output (input) to become idle reduces the efficiency of use of this waiting port.
That is, another transaction that requires this port is awaited. Instead,
If another idle port is given to another transaction, the original busy
If the destination port becomes idle and becomes busy again after a while,
The transaction will have to wait further. This delay problem is
Becomes critical if is used for large transactions.   All concurrency resolution strategies address the busy / idle status of individual ports.
Request to the controller. Maintain high transaction rates
In order to do this, the state update mechanism needs to operate with a short delay.   Transaction times are short and most delays are due to busy ports
In such cases, a complete non-blocking network topology is not required and this block
If the king probability is small and the delay is not too large, or
There is no problem as long as no extra unattainable connection requests are made.   A broadcast (one to multiple) connection is a required network capability. However, the network is broadcast
Even if you support traffic, the concurrency problem is (because this involves multiple ports).
More serious), if it is not handled without disrupting other traffic
No. Therefore, waiting for all destination ports to become idle,
The simple strategy of sending at the same time seems inappropriate.   While such special demands exist for MAN networks, MANS is not
Meets general requirements for networks. Initial costs are reasonable and the network is
It is possible to grow without confusion. This topology is for this organization
as well as It is inherently efficient in the use of circuit boards. Finally, operational availability questions
Issues such as reliability, failure tolerance, failure group size, and diagnostic and repair
The simplicity is also satisfactory.3.2. General approach: Distributed control circuit switching network   In this section, the basic approaches used within the MANS
Is explained. More specifically, large networks operate independently of each other in parallel
It will be described in terms of the means run by the controllers of the group. This variance
The control mechanism is described in the context of a two-stage network, but this approach can be used in multiple stages.
It is also possible to expand to a network with. In section 3.3, MAN
Specific design details for the embodiment will be described.   The main advantage of the approach of the present invention is that multiple network controls have only local information.
To operate independently of each other. These controls are
Throughput (measured in transactions) to avoid the burden of interrogation and response
Is Will be up. Also, since the number of sequential control steps is minimized,
The delay in setting or disconnection is reduced. All of these involve individual network organization
It uses global static information such as the internal connection pattern of the data network 120,
Its own control using only local dynamic (network state) data
This is achieved by splitting into separate subsets that are controlled only. Individual
Each controller is only interested in connection requests that use the part of the network for which it is responsible
And process this, and monitor the status of only this part.3.2.1 Division of 2-stage network   The three input switches IS1 (101), IS2 (102),
And IS3 (103), and three output switches OS1 (104), OS2 (
105) and OS3 (106).
These tissues can be subdivided into two separate subsets. Individual sub
The set is In addition to any tissue in any second stage switch (OSx), the second stage switch
Organization in the first stage switch (ISy) that connects to the link extending
Cross point). For example, in FIG. 6, related to OS1 (104)
Sub-set, ie a subset is a single dotted line surrounding the cross-point in OS1;
And three individual cross-points of the first stage switches (101, 102, 103).
(These crosspoints connect that link to OS1)
Is shown.   Now consider one controller for this subset of the network. This
It is responsible for connecting from any entry to any exit on OS1. this
The controller indicates busy / idle status for the crosspoint it controls
Hold. This information is clear enough to tell if a connection is possible or not.
You. For example, you want to connect one entrance on IS1 to one exit on OS1
And Where this request is at this entrance , Which is assumed to be idle. This output is idle
Is determined from the exit busy / idle memory or the exit in OS1.
Three cross-point states of mouth (all three must be idle)
Is determined from Next, the state of the link between IS1 and OS1 is checked.
This link is on both ends of the link connecting the link to the remaining two entrances and exits
Are idle when all the two cross points are idle. If this
If the mouth, exit, and link are all idle, each in IS1 and OS1
It is possible to close the one crosspoint and set the required connection.   This operation is handled independently of the operation in other sub-connections of the network. The reason is this net
Has only two stages, so that these entrance switches have their second switch
This is because it can be subdivided according to the link to h. In theory, this approach is all
Applies to the two-stage net. However, the effectiveness of this scheme Depends on the blocking properties of the network. The network shown in FIG.
It can happen too often, but this is at most any entrance switch.
Only one inlet on the switch can be connected to one outlet on any second stage switch
This is because   G. FIG. W. Richards (G.W.Richards) et al.IEEE transactions
Transactions on Communications, V
. COM-33, a paper published in the October 1985 issue [Two-step rearrangeable broadcast communication
(A Two-Stage Rearrangable Broadcast Switching Network)
This type of two-stage network, hereinafter referred to as Richards network,
Titles to multiple APIs distributed on different entrance switches.
Avoid by wiring to appearances. This distributed control
The MAN scheme allows the MAN to broadcast and rearrange this Richards network function.
, But not on Richards net Operate.3.2.2 Control network 3.2.2.1 Function   In a MAN, a request for a connection is made up of multiple inputs, actually MINTs.
Comes from central control 20. These requests are sent via the control network (CNet).
Must be distributed to the appropriate switch controller. In FIG.
DNet 120 and control CNet 13 for circuit switched transactions
Both 0s are shown. This DNet is a two-stage rearrangeable non-blocking liquid crystal.
Network. Each of the switches 121 and 123 is a
Troller (XPC) 122, 124, which is a specific entrance on the switch
Receives a command requesting that the
close. The XPCs (121, 123) of the first and second stages are each 1SC (one stage).
Controller) and 2SC (two-stage controller).   The right side of CNet was divided by the second stage exit switch as described above
64 M that correspond to and control the 64 separate subsets of DNet
An ANS controller 140 (MANSC) exists. These controllers and these
Is located in a layer above DNet and is not inside this data organization,
Single controller for applications where transaction throughput is not critical
Can be replaced by3.2.2.2 Structure   The CNet shown in FIG. 7 has special properties. This is three similar parts,
130, 134, and 135, which are MINT to MANSC
Message flow to MANSC, order flow from MANSC to XPC, and MAN
Positive acknowledgment (ACK) or negative acknowledgment (NA) from SC to MINT
This corresponds to the flow of K). Each of the networks 130, 134 and 135 is
It is a split switch. One bus 132, control to the destination or from the source
Roll de Interface 133 for storing data and one bus arbitration
A controller (BAC) 131 is included. Is this bus arbitration controller an input
Control the gating of control data to this bus. Address of destination
Select the output to which the bus should be gated. The output is the controller (network 1)
30: one MANSC 140) or interface (networks 131 and 132)
, An interface similar to the interface 133). Request input and
And ACK / NAK responses are controlled by the control data concentrator and distributor 136, 138.
Collected by. Each control data concentrator has four MINTs.
Alternatively, data is collected from now on. Control data concentrator and minute
The distributor simply buffers data from or to the MINT. I in CNet
Interface 133 is for statistical demultiplexing of control messages
And multiplexing (steering and merging). In DNet
Any of The interconnection made by the bus for the request message is within CNet
Note that it is the same as the required interconnect.3.2.3 Connection request scenario   The connection request scenario is a message input link from one of the data concentrators 136.
One connection request message in the form of a multiplexed data stream over one of
It starts by reaching the left of CNet130. This request should be connected
Includes the entrance and exit of DNet 120. In CNet 130, this message
The connection is led to the appropriate link 139 on the right side of CNet according to the exit to be connected.
You. Link 139 is uniquely associated with a particular second stage switch, and therefore also
Related to the MANS controller 140.   This MANSC looks up static global directories (eg, ROM)
To find which first stage switch carries the required entrance. This is other
Checks dynamic local data independently of the MANSC of the
Idol And any link from the corresponding first stage switch is idle
Find out what. If these required resources are idle, MANSC will have one
To the second stage exit switch of its own
The order is sent via the network 134 to the corresponding first stage switch.   This approach is very expensive for several reasons:
Achieve throughput. First, all network controllers must be able to communicate with each other in parallel.
Work independently and wait for each other's data or go-aheads.
There is no need. Individual controllers are dedicated only to the requests for which they are responsible,
No time is wasted on other messages. The operation of each controller is
A qualitatively sequential and independent function, and therefore one or more ongoing requests at a time
And can be pipelined.   The above scenario is not the only possibility. A possible variant is the broadcast
Into Two Poi Entry entry, exit-to-entry oriented connection request, rearrangement vs. blocking
Acknowledgment operation and selection of a block-to-busy connection request. These choices are MA
N has already been resolved. However, all of these options are provided
By simply changing the logic in the MANSC with the control topology,
Can be handled.3.2.4 Multi-stage network   This control structure can be extended to a multi-stage Richards network. here
Thus, the switches in any stage are implemented repeatedly as a two-stage network. Resulting C
In Net, connection requests pass sequentially through the S-1 controllers in the S-stage network.
I do. Again, the controller is responsible for a separate subset of the network.
Operates independently and retains high throughput capability.3.3 Specific design for MAN   This section discusses the system attributes that guide the design of MANS.
State. Next, the data and control network will be explained. In addition, design trade-offs in which the function of the MANS controller has a significant effect on performance
It is explained including.3.3.1 System attributes 3.3.1.1 External and internal interfaces   FIG. 7 shows a DNet 121 with 1024 incoming ILs and 1024 outgoing ILs.
Three controls, each with 64 incoming and 64 outgoing message links
Consists of CNet 22 including message networks 130, 133 and 134
Illustrate a typical full grown MANS. IL consists of 256 groups
Are divided into four groups for MINTs. This DNet has 64 first
It is a two-stage network consisting of a stage switch 121 and 64 second stage switches 123.
. Each switch includes one XPC 122, which is commanded and cross-ported.
Open and close the point. For each of the 64 second stages 123 of the DNet, its
One with a dedicated control link to the XPC 124 in a two-stage switch
Related to MANSC 140 exists.   Each control link and status link has four MINTs to the left of CNet
4: 1 control data concentrator on switch plane from right to right and right to left
And via the distributors 136, 138, which are also CNet2
Part 2 is constituted. These are remote concentrators within each 4-MINT group.
Or these related 1:64 CNets 130, 135 columns
Done. In the embodiment shown here, they are part of CNet
. CNet's third 64x64 plane 134 provides 64
Dedicated right-to-left interface with one link to each of the ISCs 122
Source 133 is provided. Each MINT11 is MAN through its four IL12s.
S10, its request signal to the control data concentrator 136, and the control data concentrator 1
It interfaces the acknowledgment signal received from 38.   Alternatively, each CNet may be This concentrator has 256 ports instead of 64 ports
It can be omitted.3.3.1.2 Size   The MANS diagram shown in FIG. 7 covers up to 20,000 data traffics.
Represents the network required to switch. Each NIX has between 10 and 20 EUs
Designed to concentrate S traffic on one 150 Mb / s XL
Thus, about 1000 (1024 when rounded down to binary) XL are provided. Individual
Each MINT processes 4 XLs for a total of 256 MINTs. individual
Also processes four ILs, each of which is a DNS of the MANS.
It has an input to the t-part and one output termination. This data network is thus
It has 4 inputs and 1024 outputs. The problem with the size of the internal DNet link is
This will be described later in detail.   Due to failure group size and other issues, the individual first stage switches 12
A DNet with 32 input links on 1 is considered valid You. Each of these links is connected to two switches 121. Individual of DNet
There are 16 outputs on the second stage switch 123. That is, the second stage
One switch is provided with one CNet, which includes 64 individual ties.
Switch and 64 MANSCs 140 are included.3.3.1.3 Traffic and integration   The size of the "natural" EUS transaction of data to be switched is a number
It changes in several orders of magnitude from a 100-bit SUWU to an LUWU of 1 megabit or more.
Move. As described in section 2.1.1, the MAN has a large EUS
Transactions can be network transactions or packets of up to several thousand bits each.
Disassembly into pieces. However, the MANS has one connection (and disconnection) request per request.
Switch translator defined as a burst of data passing through a MANS connection
Handle the action. The size of the switch transaction is explained below.
One SUWU by reason To several LUWUs (a plurality of packets). Section 3
In the remainder of the “transaction”, unless otherwise stated
, “Switch transaction”.   Transaction through for any total data rate through MANS
Put rate (transactions / sec) is inversely proportional to transaction size
I do. Therefore, the smaller the transaction size, the more this data rate is maintained.
Requires high transaction throughput. This throughput
Is the individual throughput of MANSC (the delay of MANSC connection / disconnection processing is not
Effective IL bandwidth) and concurrency resolution
(Time to wait for a busy exit). Individual MANSC transactions
Overhead per transaction, of course, independent of transaction size
I am in charge.   Big transactions, transactions While reducing throughput requirements, these can be extended for longer periods of time in the exit and fabric pathways.
Holding th) introduces a lot of delay to other transactions.
Small transactions reduce blocking and concurrency delays, while large transactions
Transactions reduce the workload of MANSC and MINT, and DNet
Somewhere, a compromise is needed to improve the impact coefficient of the vehicle. The answer to this is
, Its transaction so that optimum performance is achieved under varying loads of MAN.
The ability to dynamically adjust the application size.   The DNet is large enough to handle the given load, and
The limiting factor is the throughput of the exchange control complex (SCC). Light traffic
Switch state, switch transactions are short and mostly single
It consists of SUWU and packet. As traffic levels increase, transaction
Rates also increase. As we approach the SCC transaction rate limit,
, Transaction size is transac Option rate is kept slightly below the point where the SCC overloads
So that it is dynamically increased. This is a consolidation control strategy.
ol strategy). Under this strategy, individual M
INT: all existing SUWUs and packets destined for any destination
Even if each burst contains part of several EUS transactions or
Always transmitted as one switch transaction, even if it includes the whole
I do. If the traffic further increases, the size of the transaction will increase
But the number does not increase. In this way, the organization and IL utilization efficiency increase with load.
On the other hand, the workload of the SCC increases only slightly. Section 3.3.3.2.2.1
Describes a feedback mechanism that manages the transaction size.3.3.1.4 Performance target   However, MAN data throughput is limited by individual SCC controls.
Element extremely high Depends on performance. For example, each XPC 122, 124 in the data switch
Commanded to set up and tear down at least 67,000 connections per second
. Obviously, each request must be processed within a few microseconds at most.
No.   Similarly, the functions of the MANSC must be performed at a high speed. These
Assuming that the steps are pipelined, the sum of the step processing times is
It contributes to connection and disconnection delays, and the maximum of these step times is
Gives a limit on loops. For this reason, this maximum and the sum are
The goal is to keep seconds and tens of microseconds.   Solving the concurrency problem must also be quick and efficient. Destination terminal busy
/ Idle state is determined within about 6 microseconds and the control strategy is MAN
The SC must avoid imposing unrealizable connection requests.
No.   The last performance issue concerns CNet itself. this The network and this access link keep the control message transmission time short,
Run at low occupancy to minimize contention delay from statistical multiplexing
Running at high speeds (probably at least 10 Mb / s) is required.3.3.2 Data Network (DNet)   This DNet is a Richards two-stage rearrangeable non-blocking broadcast network.
. This topology is not for this simultaneous communication capability;
It is selected because it can be divided into separate subsets for control.3.3.2.1 Design parameters   The ability of the Richards network is a first stage with different entrances according to one finite pattern.
Gained from assigning multiple appearances on a given switch
It is. Specific allocation pattern selected, multiple appearances per entrance
The number of doors, the total number of entrances, and the number of links between the first and second stage switches.
The second allowed to rearrange without blocking the network The maximum number of outlets per stage switch is determined.   The DNet shown in FIG. 7 has two Apiarans each on the first stage switch.
It has 1024 entrances with shrinkage. Two between each first and second stage switch
Link exists. These parameters are the pattern for distributing these entrances.
Together with the switch, 16 outlets were provided per second stage switch.
Can be rearranged so that this network does not block broadcasts.
Secured.   MAN does not use broadcast or reordering, so failure groups or other
These parameters, which cannot be justified due to problems, change as experience gains
It is possible. For example, ensure that 32 failure group sizes can be tolerated.
If approved, each second stage switch can have 32 outputs, and
Thus, the number of switches in the second stage can be reduced by half. You can make this change
Whether or not it depends on the ability of the SCC control element to handle individual traffic
Dependent. In addition to this, the probability of blocking increases, which increases network performance.
It must be confirmed that it will not fall off.   This network comprises 64 first stage switches 121 and 64 second stage switches.
It has 123. Each entrance has two appearances, the first and second stages
Because there are two links between the switches, each first stage switch has 32
With 128 inlets and 128 outlets, each second stage having 128 inlets and
It has 16 outlets.3.3.2.2 Operation   Each entrance has two appearances, and each first and second stage
Since there are two links between switches, any exit switch will have four links
Can be accessed at any one of the entrances. Entrance to link
Is algorithmic and therefore can be calculated or read from a table.
Can also be issued. The path hunt is simply an idle (exist
The link). You.   If none of these four links are idle, you will need to set up a connection later.
Attempt again by the same MINT. Alternatively, existing connections
May be rearranged to remove the blocking condition, but this is
This is a simple procedure for a chard network. However, the root of the connection in the middle stream
Any changes in the ringing will cause the exit circuitry to exceed the phase grid's ability to recover the phase and clock.
May be introduced. Therefore, in this circuit, a switch capable of rearranging MANS is used.
It is better not to run as a switch.   Each switch in the DNet connects one XPC 122, 124 on CNet.
This is a message that tells the MANSC which crosspoint to operate.
Receive a message. Some of these controls perform high-level logic.
Absent.3.3.3 Functions of control network and MANS controller 3.3.3.1 Control network (CNet)   The CNets 130, 134, 135 described briefly above are MINT, MAN
SC and 1SC are interconnected. These are three types of messages:
, Connection / disconnection order from MINT to MANSC using block 130,
Crosspoint order from MANSC to 1SC using lock 134
, ACK and NA from MANSC to MINT using block 135
Carry K. CNet shown in FIG. 7 has three corresponding planes:
With an option. 124 links of private MANS140-2SC are shown
, Are not considered part of CNet because they do not require exchange.   In this embodiment, the 256 MINTs are grouped into CNet in four groups.
And therefore 64 input paths to the network and 64 output paths from the network.
Roads exist. Bus elements in the control network manage the merging and routing of message streams.
Perform. Exit to be connected or disconnected in request message from MINT
Contains port ID It is. Since the MANSC is associated one-to-one with the second stage switch, this exit designation
Identifies the correct MANSC to which the message should be sent.   The MANSC sends a positive acknowledgment (ACK), a negative acknowledgment (NAK), and 1
The SC command message is sent from the right-to-left portion of CNet (blocks 134, 1
Carry through 35). This message also specifies the MI
Heading information for routing to NT and 1SC is included.   CNet and this message demand great technical challenges. Competition in CNet
The joint problem reflects the overall contention problem of MANS and calls for a unique concurrency solution. this
Are evident from the control network shown in FIG. One out of four lines
The control data concentrator 136 to the interface allows one or more messages to be
If you try to reach it, you will create a conflict. The data concentrator 136 has four connected
MINT has memory for one request from each of MINT, and MINT has a series of requests.
Is MINT The previous request is far enough away to be passed by the concentrator before the next request arrives.
And send it to you. MINT responds to requests within a predetermined period.
Timeout if no acknowledgment is received. Alternatively, the control data collection
Receiver 136 simply translates any request received on any entry to this exit into "O
R ”processing, ignore error requests, do not send acknowledgments, and time out.
Can also be.   Functionally required within blocks 130, 134, 135 include:
Micro-L dedicated to always small fixed length packets, low contention and low delay
AN. Ring nets are easy to interconnect, grow gracefully, and simply
Allows pure tokenless add / drop protocols, but these
It is not suitable for densely packed nodes and has a long end-to-end delay.   Even the longest message (MINT connection order) is less than 32 bits
The parallel bus CNe 132 can send one complete message in one cycle
Functions as t. The arbitration controller 131 is responsible for processing this bus contention.
Thus, the conflict for the receiver is automatically resolved. Bus elements are heavy for reliability purposes
It is duplicated (not shown).3.3.3.2 Operation of MAN switch controller (MANSC)   FIGS. 8 and 9 show flow charts of the high level functions of the MANSC. Individual MAN
The message to SC 140 includes a connect / disconnect bit, SUWU / packet bit.
And the ID of the MANS input and output ports involved.3.3.3.1.2.1 Request Queue; Integration (Intake Section, Fig. 8)   The speed of messages arriving at an individual MANSC 140 is determined by this message
MANSC may be the entry point to the message
Provide a queue. Connect and disconnect requests are handled separately. Connection requests are
request Will not be placed in the queue unless you have been exited.   Separate queues 150, 152 are provided for priority and normal packet connection messages
Priority packets are given higher priority. From the normal packet queue 152
Are processed only when the priority queue 150 is empty. This is the priority packet
Process delay at the expense of legacy packet processing delay. However, priority
Traffic is usually not expected to be high enough to cause large packet delays.
It is. Nevertheless, high-priority, high-volume data transactions are higher
Users tend to tolerate the delay more than in the case of the transaction. Ma
Also, if a packet is one of many pieces of a LUWU, the delay of any packet
Since the end-to-end LUWU delay depends only on the last packet,
It is not expected to significantly affect the results.   Priority and normal packet queuing is short, short-term random variations in message arrival.
Cover only movement The porpose is to do. If this short-term speed of arrival exceeds the processing speed of MANSC
The normal packet queue, and possibly the priority queue, will overflow.
You. In such a case, a control negative acknowledgment (CNAK) is returned to the request MINT.
To indicate that the MANSC is overloaded. This is not a catastrophe
However, when traffic increases due to the feedback mechanism in the integration strategy,
Only the size of the switch transaction is increased. Each MINT is
All existing packets destined for a certain DNet egress are
To the action. Thus, as a result of the connection request by MINT, CNA
If a K is received, the next request for the same destination will have a L
If more UWU packets arrive, more data will be carried during this connection.
It will be. Since the last packet of the LUWU may not be affected,
In this case, the LUWU transmission delay is not always increased. This scheme is MA
NSC processing capacity Dynamically increase effective packet (transaction) size to help power
Add.   To reduce the probability that a priority CNAK will be sent due to a random burst request.
For this reason, the priority queue is usually made longer than the packet queue. The priority packet is
Less likely to benefit from consolidation than packets recombined into LUWU
Support this separate high priority queue. Integrate more packets into MINT
The packet queue to be shorter than it should be
Can be. The simulation results show a priority queue with four required capabilities and
Indicates that a normal queue with eight required capabilities is appropriate. Rhino in both queues
Affects the performance of the system and fine-tunes based on the actual experience of the system
be able to.   The priority is determined by the priority index within the type of service index 623 (FIG. 20).
It is. Voice packets are given priority because they require a small delay.
All single packet transactions (SUWU) can be given priority. Charges for high priority services
Can be increased, so the user has to
Requesting high-priority services is likely to be reluctant.3.3.3.2.2.2 Busy / idle check   When the connection request first arrives at the MANSC, this is
Although detected, this test determines this from the disconnect request. Destination Exit Business
-/ Idle status is checked (test 154). If the destination is busy
Returns a busy negative acknowledgment (BMAK) to the request MINT (act 156);
The requesting MINT will try again later. Test 158 returns the corresponding queue (priority
Or normal packet). This queue is tested to see if it is full
(160, 162). If the designated queue is full, CNAK (control negation)
The target acknowledgment is returned (act 164). Otherwise, the request is queued 150 or 152 while simultaneously capturing the destination (marked busy and
(Operation 166 or 167). Overwork MAN
SC also returns BNAK and both BNAK and CNAK transact through integration.
Note that it tends to increase the action size.   This busy / idle check and BNAK address the issue of concurrency. This
The price paid for this approach is that the IL from MINT to MANS is MIN
T receives an ACK or BNAK after T issues a connection request for that IL.
You cannot use it until you trust. Also, when the load on the MANS is large
Below, CNet is jammed by BANK and failed requests. busy
/ Idle check should not reduce connection request throughput and IL usage
You must perform fast enough. This is a busy test before queuing
Explain the need for In addition, egress concurrency using separate hardware May need to be tested in advance. This procedure is MANSC
And release CNet from repeated BNAK requests and increase the success factor throughput
Was. Also, MANS saturates at a higher percentage of its theoretical overall bandwidth.
To be able to3.3.3.3 Route Hunt-MANSC Service Section (Fig. 9
)   Priority block 168 gives the highest priority to requests from disconnect queue 170,
The request from queue 150 is given a lower priority and packet queue 15
Give the request from 2 the lowest priority. Connection request has priority or normal packet wait
When unloaded from the queue, this requested exit port has already been seized.
(Operation 166 or 167), and the MANSC passes through the DNet.
Hunt the road. This simply looks at the two entrances where the IL is connected first (
Operation 172), finds four links with access to the incoming IL, and
Check status (test 174). If all four are busy, tissue block N
AK (Organization NAK) or Organization Block Negative Reception Communication (FNAK)
Returned to INT, the request MINT will retry the request later (act 178). Also capture
The captured destination exit is released (marked idle) (operation 176). F
NAK is expected to be rare.   If all four links are not busy, one of the idles is selected, and
First, the first stage entrance, then one link is captured (act 180) and both
Are marked as busy (operation 182). Next, MANSC has its dedicated control
Cross-point to the XPC in its associated second stage switch using
A connection order is sent (operation 188). This makes the selected link its exit
Connecting. At the same time, another crosspoint order (CN from right to left)
request to connect the link to the ingress port (via the et plane 134)
Sent to 1SC (Operation 186). When the order reaches 1 SC (test 190), A
CK is returned to the originating MINT (operation 192).3.3.3.3 cutting   In order to release network resources as quickly as possible, the disconnection element is
Is processed with priority. These are 16 words so that they do not overflow
It has a separate queue 170 created by length (equal to the number of exits). The disconnect request is MI
Upon receiving a request from the NT, a test 153 for determining a connection request and a disconnection request is performed.
Detected. The exit is released and the request is placed in the disconnect queue 170 (
Operation 193). Now, if the new connection request for this same exit is
Acceptable even if not physically disconnected. This has a high priority
The disconnect request disconnects the switch connection before a new request attempts to reconnect this exit.
I do. Once in the queue, the disconnect request is always executed. Consumed consumption
Only the exit ID is needed to identify the continuation. MANSC is this Recall the link and crosspoint selection of the connection from local memory (behavior
195), mark these links as idle (operation 196) and release them.
Send two XPC orders to send (acts 186 and 188). Then,
Strike 190 controls the waiting for an acknowledgment from the first stage controller,
An ACK is sent to MINT (operation 192). If there is no record of this connection,
MANSC returns "sanitary NAK". The MANSC provides phase matching and scanning at its exit.
Detects the state from the scramble circuit (PASC) 290, and data transmission occurs.
Check if it is.3.3.3.3.2.5 Parallel pipe lining   Except for resource capture and release, the above steps for a single request are the same.
It is independent of other request steps within one MANSC and therefore
Pipelined to improve puts. Through parallel operation
Greater power is achieved. In other words, route han Starts at the same time as the busy / idle check. Transaction speed
Depends on the longest step in a pipelined process, but is optional
Response time for transaction (from request to ACK or NAK)
Is the sum of the step times involved. The latter is based on parallelism.
However, it is not improved by pipeline connection.3.3.4 Error detection and diagnosis   CNet to verify all small messages and high cost for this node
Avoids hardware, message bits, and time-consuming protocols
Is done. For example, individual crosspoint orders from MANSC to XPC are
It does not require a command echo, or even an ACK returned. MANSC
External evidence shows that the sage arrived without loss and was treated normally.
Assume until it reaches. Audits and cross-checks are made when reasons for doubt exist.
Only be activated. End users, NIM and MINT are Or directly detect defects in the control complex and participate in the MANS ports involved
Identify a subset of The diagnostic task then tracks the problem for repair
You.   Once a part of the MANS is suspected, a temporary audit mode will identify the suspect.
Turned on to catch. For the alleged 1SC and MANSC, this
Mode requires the use of the command ACK and echo. Special messages, examples
For example, a crosspoint audit message is passed in CNet. This is the user
  It should be performed in conditions carrying a light load of traffic.   Before embarking on these internal self-tests (or completely removing them)
, MAN uses MINT, IL, and MIN to identify the faulty circuit.
Perform the test above. For example, 75% of SUWU sent from any IL
When passing to the exit, the two links from one of the two first stages of the IL
One can conclude that one is defective. (This test is deterministic It is not performed under load so that the MANSC does not always select the same link.
Must be). Further testing can identify the fault link
It is possible. However, multiple MINTs have been tested, all of which
If not, the exit shall be “out of service” for all MINTs.
And suspicion is narrowed down to the second stage and the MANSC.
If the other outlet works, then the fault is in the second tier of tissue. These tests are
Use the status read from the 16 PASC of each MAMSC.   To adjust the independent MINT and MIN to run these tests,
Central intelligence with low bandwidth message links to all MINTs and MINs
Required. Given the connection between MINTs (see FIG. 15), the required c
Any MINT with hardware can perform the diagnostic task. Where is NIM
Needs involvement even if test SUWU reaches its destination
Becomes Of course, all of the work MINT The NIM can exchange messages with all other such NIMs.3.4 MAN switch controller   FIG. 25 is a diagram of the MANSC 140. This is to set up the circuit connection
Or a unit that sends a control command to the data network 120 to disconnect it.
You. It receives orders from control network 130 via link 139,
In addition, the control network 136 is sent to the MINT11 requesting both positive and negative acknowledgments.
Sent through. This also causes instructions to be sent to the first stage switch controller in the first stage.
Via a control network 134 to the switch controller 122 of the
Instructs the second stage controller 124 directly associated with a particular MANSC 140
Send.   Input is received at entrance 139 at request receiving port 1402. These are accepted
Processed to determine if the exit requested by control 1404 is not busy
It is. Exit memory 1406 is exit busy / idle for which MANSC 140 is responsible
index including. If the exit is idle, the two waits described earlier in connection with FIG.
Placed in one of the matrices 150 and 152. When the request is for disconnection
, The request is placed in a disconnect queue 170. Exit map 1406 disconnected
Updated to mark exits as idle. Reception / response unit 1408 required
The connection request was made to a busy exit when there was an error in receiving the request.
Negative, or if the corresponding queue 150 or 152 is full
Send receipt notification. The reception response is transmitted via the control network 135 to the request MINT11.
Via the distributor 138. All of these actions are performed by the reception control 14
04 under control.   The service control 1420 controls the setting of a route in the data network 120,
There is also a route available in the data network between the request input link and the vacant output link.
If the information does not exist, the exit memory 1406 is updated. Receipt control is also connected
Exit memory 1406 already in queue for request Updates an existing request to block another request for the same output link.   Service control 1420 has three queues 150, 152, and 170
Examine the request within. Disconnect requests are always given the highest priority. For disconnection request
Link memory 1424 and path memory 1426
Examine to find out what to do. The instruction to idle these links is
Sent to first stage switch from first stage switch order port 1428
And instructions to the second stage switch are sent to the second stage switch order port 1
Sent from 430. For a disconnection request, the static map 1422 displays the request input
Link should be used to route the link to the required output link.
Examine to find out. Next, the link corresponding to the link map 1424 is used.
Are checked to see if they can be used, and if available they are marked as busy
Is done. Route memory 1426 is updated to indicate that this route has been established.
, This allows the corresponding link to be idle when a disconnection order comes later.
It becomes possible. All of these actions are controlled by the service control 1420.
Performed below.   Controllers 1420 and 1404 can be separate controllers, even if they are single controllers.
It can be a controller, it can be program controlled, or it can be
It can also be controlled by logic. These controllers require high throughput.
Requires very high-speed operation, so hard-wired controllers
Is preferred.3.5 Control network   Control message network 130 (FIG. 7) outputs 137 to data concentrator 1
36, and output these signals representing connection or disconnection requests to the MAN switch.
Controller 140. The output of the concentrator 136 is temporarily stored in the transmission register 133.
Is stored. The bus access controller 131 sends these transmission registers 13
Poll 3 to see if it has a request to be sent. These points The request is then placed on bus 132, the output of which is temporarily stored in intermediate register 141.
To be placed. The bus access controller 131 then outputs from the register 141.
Force is applied to the appropriate one of the MAN switch controllers 140 via link 139
The output of the register 141 on the bus 142 connected to the link 139
I will send you. This operation is achieved by three phases. First phase
, The output of register 133 is placed on bus 132 from which register 1
Gated to 41. In the second phase, the output of register 141
42 and delivered to the MAN switch controller 140. 3rd
In the phase, the MAN switch controller controls the transmission register 133.
Notifies whether the roller has received the request. If received, send register 13
3 can accept a new input from the control data concentrator 136. So
If not, the transmission register 133 holds the same request data and the bus access
  controller 131 will try again later. These three phases correspond to three separate requests
Can occur simultaneously. Control networks 134 and 135 are
It operates in a similar manner.3.6 Summary   High bandwidth and transaction throughput requirements for MANS
A structure for satisfying has been described. The data switch organization has
Pipelined distributed switch control with parallel locking probabilities
This is a two-stage Richards network selected because it allows for a complex (SCC).
This SCC is the XPC in all the first and second stage switches, including the individual second stage.
Intelligent controller MANSC and control fragment
Contains CNet linked to MINT.4. Memory and interface modules   The memory and interface module (MINT) is
Interface, buffer memory, routing and Link for the protocol and the data collected
To provide a transmitter for sending to the MAN switch through. In the design of the description
Each MINT handles four Network Interface Modules (NIMs)
, With four links to the switch. MINT is a data exchange module.4.1 Basic functions   The basic function of MINT is to provide:   1． Fiber optic receiver and link protocol for each NIM
Dora.   2． Link handler and transmitter for each link to the switch.   3． Buffer memory to accommodate packets waiting to be transmitted across the switch
Mori.   Four. Input to the controller for the switch that commands the setting and disconnection of the network route
Face.   Five. Address translation, routing, efficient use of switches, collected packets
Orderly transmission of And controls for buffer memory management.   6． Interface for operation, supervision and maintenance of the entire system.   7． Control channels to individual NIMs for operation, supervision, and maintenance functions.4.2 Data flow   To understand the description of the individual functional units that make up MINT, first,
A basic understanding of the general flow of data and controls is required. Figure 10
1 shows an overall image of MINT. Data is stored in MINT at high speed (10
(0-150 Mbit / s) carried by data channel 3. This data
, Each of the order of 8 kilobits long has its own heading containing routing information
One packet format. This hardware has 512 bit increments
To accommodate packet sizes up to 128 kilobits. However, small
The packet size reduces throughput due to per-packet processing requirements. You. A large maximum packet size is a transaction that is smaller than the maximum size packet.
Wastes memory for the application. Link to external link handler 16 (XLH)
Terminate, but this is when this puts the entire packet in its buffer memory.
, Keep a copy of the required heading section. This heading information is then stored in the buffer note
It is passed to the central control 20 along with the readdress and length. Central control
The role determines the destination NIM from the address and transmits this block to this same destination.
Add to the list of blocks waiting to be sent (if any). Central control also
Connect request to this switch controller if there are no outstanding requests already
Send. Central control satisfied connection request from switch controller
Upon receipt of the acknowledgment, the central control
Is sent to the corresponding external link handler 17 (ILH). ILH is stored
Read data from memory and read it at high speed (probably the same speed as the incoming link)
At) MAN Switch and the MAN switch points it to the destination. This block is transmitted
When ILH is used as a central control, this block can be used by XLH
Notify that a link is added to the available free block.4.3 Memory module   The buffer memory 18 (FIG. 4) of the MINT 11 satisfies the following three requirements.
Let it.   1． The amount of memory is set by the switch to collect the data (for all destinations)
Provide enough buffer space to wait and hold the.   2． Memory bandwidth is available for all eight links (four receive and four transmit links).
H) is sufficient to support the above simultaneous operations.   3． Memory accesses affect the effect of data to and from the link handler.
Provides a streamlined flow.4.3.1 Organization   Due to the amount of memory required (megabytes), traditional high-density dynamic random access
Access memory It is necessary to employ (DRAM) parts. Therefore, high bandwidth is not
Can only be achieved by widening the The memory therefore has 16 modules.
Rules 201,. . . , 202 so that the compound 512-bit word is
Be provided. As you can see below, memory access requires any module
Do not continue to receive requests so that there is not enough time to complete the cycle
It is organized in a synchronous manner. One M in a typical MAN application
The range of memory for INT11 is 16-64 Mbytes. This number is
Sensitive to the speed of flow control applications in bar load conditions
is there.4.3.2. Time slot allocator   The time slot allocators 203,. . . , 204 (TSA) is a conventional SRAM core
Controller and the functionality of a dedicated 8-channel DMA controller. Individual
Read / write from logic associated with data transmission ring 19 (see section 4.4)
Only Receive the request. This configuration command is a dedicated control on this same ring
Coming from a time slot.4.3.2.1 Control   From a control point of view, TSA is a set of cash registers as shown in FIG.
Looks like a star. For each XLH, the associated address register
Data 210 and a count register 211. Each ILH is also an address
It has a register 213 and a count register 214.
It has a register containing a dress 215 and a count 216, and thus
Data from memory as a continuous stream without gaps between blocks.
And enable. A dedicated set of registers 220-226 is the central control of MINT.
Note that the role section accesses any internal registers in the TSA
A specific word from the directory as it is read or written to it.
You. These registers include a data write register 220 and And data read register 221, memory address register 222, channel
Status register 223, error register 224, memory restoration row address register 2
25, and a diagnostic control register 226.4.3.2.2 Operation   In normal operation, the TSA 203 determines from the ring interface logic that
There are four order types: (1) for data received by XLH
“Write” request, (2) “read” request for ILH, (3) XLH or ILH
"New address" command issued by (4) TSA
Or only an "idle cycle" indicator telling it to perform some other special action
To receive. Each order is accompanied by an identification of the link handler involved,
For "write" and "new address" requests, 32 bits of data accompany.   For a "write operation", the TSA 203 simply associates with the indicated XLH16.
From register Using the address and data provided by the ring interface logic
Perform a memory write cycle. This then increments the address register,
Decrement und register. In this case, the count register indicates that XLH is the current block.
Safety check to provide new address before lock overflows
Used only as   For a “read” operation, the TSA 203 first
You must check if the panel is active. This is active
, The TSA uses the address from the register for this ILH17
To perform a memory read cycle, and transfer this data to the ring interface logic.
To provide. This also increments the address register and decrements the count register.
Minute. In each case, the TSA has two "tag" bits in the interface logic.
Provides (1) no data, (2) data, (3) packet
Indicates that the first word is present, or (4) the last word of the packet is present. Case (4)
Against In other words, TSA uses the ILH address register 214 and count register 213.
From the “next address” 216 and “next count” 215 registers,
These registers are loaded on the assumption that they have been loaded by the ILH. this
If they are not loaded, the TSA sets this channel to "inactive".
".   From the above description, the function of the "new address" operation can be inferred. TSA203 is phosphorus
Receive a 24-bit address, and an 8-bit count. XL
For H16, this simply loads the associated register. Place of ILH17
If so, the TSA must check if the channel is active
. If not active, normal address 214 and count 213 registers
Is loaded and the channel is marked as active. Channel is currently active
If so, replace the normal address and count register with the “next address”.
Address 216 and the "next count" 215 register are loaded.   In another embodiment, the two tag bits are stored in buffer memory 201.
,. . . , 202. As an advantage, this is the overall width of the memory (51
Allows packet sizes that are not a multiple of 2 bits). In addition to this, ILH
17 need not provide the actual length of the packet when reading it out,
This eliminates the need for control 20 to transmit this information to the ILH.4.4 Data transmission ring   The job of the data transmission ring 19 is linked to the link handlers 16 and 17 and the memory module.
Modules 201,. . . Control commands and high-speed data between
To carry. This ring is enough band for all links to run at the same time
Provide width. However, this is determined by the circuitry that connects this bandwidth to this ring.
Be careful not to be required to transmit data in high-speed bursts.
Guess. That is, a fixed time slot cycle is adopted,
Allocated to individual circuits at well separated intervals. This fixed The use of fixed cycles also requires that the source and destination addresses be carried on the ring itself.
Means no need. This means that they are correctly synchronized at any point.
This is because it can be easily determined by the determined counter.4.4.1 Electrical description   This ring is 32 data bits wide and is clocked at 24 MHz.
This bandwidth is sufficient to support data rates up to 150 Mbit / s.
Minutes. In addition to these data bits, the ring has four parity bits, 2
One tag bit, one synchronization bit identifying the start of a superframe, and
Includes one clock signal. Within the ring, remove any clocks that are differential ECLs.
Therefore, an ECL circuit which is unbalanced terminated is used for all signals. Rings
The interface logic provides TTL compatible signal levels to the connection circuit.4.4.2 Time Slot Sequence Requirements   In order to fulfill the above objectives, The cycle has the following restrictions:   1． In each complete cycle, the combination of each source and destination
There must be one unique time slot for   2． The individual connection circuits have data time slots appearing at appropriate regular intervals.
Must have Specifically, each circuit has a data time slot.
Must have a minimum period between them.   3． Individual link handlers are numerically ordered by memory module number
  Must have a time slot. (This is because the link handler is 51
This is to avoid shuffling 2-bit words).   Four. Each TSA may perform a restore cycle or other miscellaneous memory operations in the meantime.
Must have a known period in which to go.   Five. TSA in memory module is all control time slots
Must be examined between control time slots. There must also be a minimum period.4.4.3 Time slot cycle   Table I shows one data frame of a tamping cycle that meets these requirements.
Shows the game. One data frame consists of a total of 80 time slots.
Of which 64 are used for data and the remaining 16 are used for control
Is done. The table shows for each memory module TSA, in the meantime,
The data to be written into the memory is received from the individual XLHs, during which time
Requires that it supply data read from memory for each ILH
Indicates the slot to be used. Every 5th slot is a control time slot
During this time, the indicated link handler takes control order.
Broadcast to all TSAs. In this table, XLH and ILH
Numbers 0-3 are given and TSAs are given numbers 0-15. For example,
TSA0 receives data from XLH0 in time slot 0 and ILH0 You must supply data for In slot 17, TSA0 is
Perform similar operations on XLH2 and ILH2. Slot 46 is XLH1 and
Slot 63 is used for XLH3 and ILH3
Used. XLH never reads from memory, and ILH
Since no write is performed at the same time,
Reuse is allowed, thereby effectively doubling the data width of the ring.   The control time slots consist of four XLHs, four ILHs, and
And central control (CC). These nine entities are controlled
Control frames share 45 time slots.
G. 80-slot data frame and 45-slot control
Frame aligns once with a 720 time slot. This cycle is
And is marked by the superframe synchronization signal.   There are subtle synchronization conditions that must be met for ILH. One block
Term relates to where in the ring timing cycle the order was received.
And must be sent in order starting with word 0. To fulfill this requirement
In order to assist the ring interface circuit, a special "word"
0 "synchronization signal. For example, in the timing cycle of Table I,
, In time slot 24 (its control time slot)
Assume that a new address is sent by H0. Here, for ILH0
Data time slot for reading from TSA numbered 5 to 15
TSA number 0 will be assigned to this new address even if the
Be the first TSA to work on the dress (Requirement of section 4.4.2)
It is necessary to guarantee 3).   The number 720 of time slots in the superframe is 25 times the number of elements in the ring.
To exceed It is clear that logical time slots have no permanent existence. That is,
Each time slot is generated at a specific physical location on the ring,
, And returns to this position, where it disappears. Generation point is data
The time slot and the control time slot are different.                                Table I                     Ring time slot 4.4.3.1 Data Time Slot   Data time slots can be considered to occur at their own XLH
You. Data Thailand System slot carries incoming data to its assigned memory module and
At this point, it is reused to carry outgoing data to the corresponding IHL. X
Since the LH never receives information from the data time slots, the ring
Logically broken between LH and XLH (only for data time slots)
Can be regarded as having.   Two tag bits identify the contents of the data time slot as follows:
You.     11 Sky     10 Data     01 First word of packet     00 Last word of packet   This "first word of the packet" is only
Sent when sending the first word of a packet to the ILH. Of the "last word of the packet"
The indicator is only by the memory module 15, which ends the packet with ILH
Sent when sent. 4.4.3.2 Control Time Slot   The control time slot is the status of the central control 20 on the ring.
Occurs and terminates at the The link handler uses these assigned
Control slots are only used to broadcast orders to the TSA.
CCs are assigned every nine control time slots. TSA is
Receives orders from all control time slots and responds with CC control
Sent to CC on roll time slot.   The two tag bits describe the contents of the control time slot as follows:
Identify.     11 Sky     10 data (to or from CC)     01 order     00 Address and count (from link handler)4.5. External link handler   The main function of the XLH is to enter the NIM and terminate the high-speed data channel.
Data in the MINT buffer memory and data can be transferred to the destination
In this way, the information required for the central control 20 of the MINT is passed. This
In addition, the XLH is an incoming low-speed control multiplexed over optical fiber.
Terminate the tunnel. Some functions assigned to the slow control channel
Includes transmission of NIM state and control of flows in the network. XLH from NIM
Note that only the incoming fiber is terminated. Transmission to NIM is an internal link
Handled by the handler and the phase matching and scrambler circuit described below.
XLH only interfaces to the hardware of MINT Central Control 20
For this purpose, an on-board processor 268 is used. The four coming from this processor
20Mbit / s link provides connection to MINT's central control section
provide. FIG. 12 shows the entire XLH.4.5.1 Link interface   XLH is a fiber optic receiver needed to recover data from fiber
, A clock recovery circuit and a descrambler circuit. Data clock recovered
(Block 250) and the data is descrambled (block 252).
), The data is converted from serial to parallel, and a high-speed data channel
And demultiplexed to the low speed data channel (block 254). Low
The level protocol process is then described in Section 5 for this data
Performed on the high speed data channel as described (block 256). This result
As a result, a data stream consisting of only packet data is provided. Packet data
The data flow then passes through the data steer through a first in first out (FIFO) queue 258.
To the switching circuit 260. Circuit 260 sends the header to FIFO 266, and completes the
The packet is sent to the XLH ring interface 262.4.5.2 Link interface   Ring interface 262 is a link-in From the packet FIFO 258 in the interface to the buffer memory of the MINT.
Controls data transmission. It provides the following functions: 1． Establish and maintain synchronization with the ring timing cycle. 2． Link interface FIFO to applicable ring time slot
Data transmission. 3． Transmission of a new address to the memory TAS when the end of the packet is encountered.   Links ring resynchronization with 16 word timing cycles (per XLH)
Perform when processing a packet each time the interface FIFO is temporarily emptied.
Note that you must: This is the bandwidth of the ring is the transmission speed of the link
Always higher because it is higher. However, the ring and TAS do not
Designed to accommodate a laptop. Thus, resynchronization simply involves data coming
Or the time it takes for the ring cycle to return to the correct word number
It consists of marking the slot as "empty". For example, F IFO 258 emptied when word for fifth memory module needed
In order to keep the entire sequence, the next word is actually
It is necessary to ensure that they are sent to the rules.4.5.3 Control   The control section of XLH refills and receives empty blocks FIFO 270
The heading information about each of the individual packets is stored in the MINT central control 20 (
(Fig. 4).4.5.3.1 Index processing   At the same time as the packet is transmitted on the ring, the header of the packet is
FO 266, which is later read by XLH processor 268.
You. Within this heading are the statements that the central control needs for routing.
Source and destination address fields are included. In addition to this, the header checksum
Are verified to ensure that the fields have not fallen. Heading information is stored in memory
  Block descriptor (address and length) And a centralized control (Fig. 4) in one message
Sent.4.5.3.2 Interaction with Central Control   There are basically only two interactions with the MINT central control. XL
H control gets the FIFO 270 of the empty block from the memory manager
Can be sent to its destination.
Pass header information and memory block descriptors to the central control
You. The block address is then placed on ring 19 by ring interface 262
Is placed when the address is received from the control sequencer by
You. Both interactions with the central control are performed by the XLH processor 268 from the central
Carried through a link to the appropriate section of the control.4.6 Internal Link Handler   The internal link handler (ILH) (Figure 13) is a distributed link controller
This is the first part that can be considered as At any moment in time This distributed link controller passes through one particular ILH, switch organization.
One path and one particular phase matching and scrambler circuit 290 (P
ASC). PASC is described in section 6.1.
Is done. Return fiber pair from MINT to NIM through fiber 3
It is this PASC that has the actual responsibility for transmitting all optical signals. This fiber
The information transmitted through comes from the MANS10, which inputs at various times.
Received from the ILH to be transmitted to the NIM. This type of distributed link controller
This is necessary because the path lengths through the MAN switch organization are not all the same.
If the PASC does not match all of the information coming from different ILHs to the same reference clock
, NIM always fluctuates its phase and bit alignment.
And   The binding of ILH to PASC is in many respects a mirror image of XLH. ILH
Receives a list of block descriptors from the central control Read these blocks and switch the data through the serial link.
Send to Ji. When data is received from memory, the associated block descriptor is blocked.
To the central control memory manager so that the
I'm done. ILH, XLH, and ILH do not perform special heading processing and need
TAS adds additional pipeline connections to ILH depending on multiple blocks in one
The difference is that it is provided so that it can be transmitted as a stream.4.6.1 Link interface   Link interface 289 provides a serial transmitter for the data channel.
provide. Data is linked data format as described in section 5
It is transmitted in a frame synchronization format compatible with the packet. Data links
  Asynchronously from interface 280 to a speed somewhat higher than the average speed of the link
Link interface provides speed matching and frame synchronization.
And a FIFO 282 for serving. De Data received from MINT memory via data ring interface 280
And stored in the FIFO 282 and passed to the level 1 and 2 protocol handlers.
And the parallel / serial conversion in link interface 289.
The signal is sent to the MAN switch 10 through the exchanger 288.4.6.2 Ring interface   Ring interface 280 logic links from MINT buffer memory
  Controls the transmission of data to the FIFO in the interface. It has the following features
I will provide a. 1． Establish and maintain synchronization with the ring timing cycle. 2． From the ring in the corresponding ring time slot to the link interface
Transmission of data to the source FIFO. 3． Control when the last word of a packet (memory block) is received
Contact section. Four. The last word of the packet is received, and the state of FIFO 282 indicates this new packet.
TAS 203 in a case where the data does not cause overflow
,. . . , 204 (FIG. 10) new address and count (if any)
) Sending.   Unlike XLH, ILH receives data words in sequence in TSA and
It relies on the TSA to ensure that there are no gaps in the lock. Therefore,
Retention of word synchronization can in this case simply result in unexpected empty data time slots.
Consists of searching.4.6.3 Control   The control part controlled by the sequencer 283 of ILH is a ring.
The central interface is connected to the interface via the processor link interface 284.
Block received from the controller and stored in the address FIFO 285 from there.
Processor link interface to the central control.
Block through notes Notify that it has been removed from the library and send the last block to the central control 20.
Is responsible for notifying that the transmission has been completed.4.6.3.1 Interaction with Central Control   There are basically only the following three interactions with the MINT central control:
You. 1． Receiving a list of block descriptors. 2． Notification of blocks fetched from memory to the memory manager. 3． All blocks have been transmitted to the switch request queue manager
Notifications.   In the design described here, all of these interactions correspond to the central control.
This is accomplished through a transputer link to the relevant section.4.6.3.2 Dialogue with TSA   Like the XLH, the ILH uses its control time slot to block
Sends the TACK descriptor (address and length) to the TSA. However, if TSA is ILH
As soon as they receive the descriptor, they read the block from memory and
Up Start the operation to put the data in. The length field from the ILH is important,
Determine the number of words that can be read by each TSA before moving on to the task. T
The SA also provides individual ILHs so that a series of blocks can be transmitted without gaps.
Provides a register for holding the next address and length. Flow control
Roles are the responsibility of the ILH, except that new descriptors are subject to reframing time and
There is enough room in the packet FIFO 282 to compensate for differences in
Should not be sent until they are sent.4.7 MINT Central Control   FIG. 14 is a block diagram of the MINT central control 20. This central console
Trolls consist of four XLH16s of MINT, four ILH17s of MINT,
Switch (FIG. 7) data concentrator 136 and distributor 238, and
It is connected to the OA & M central control 352 shown in FIG. First, center
The relationship between the control 20 and other units will be described.   MINT Central Control communicates with XLH16 and enters data via XLH
The memory block address used to store the
provide. XLH16 communicates with the MINT central control to store in the MINT memory
The header of the packet to be stored in, and where this packet should be stored
Provide an address without indicating Memory management of MINT Central Control 20
The memory communicates with the ILH 17 to store the memory stored in those memory blocks.
Receiving information indicating that the memory delivered to
The memory released by is reused.   The first network unit that queue manager 311 arrives for a particular NIM
Has been placed in the switch unit queue 314,
Queue 316 for the possible destination NIMs, but the queue manager
Reference numeral 311 denotes a switch setting control 313 to the MAN switch 10 to the NIM.
Request to connect within Send a signal. This request is sent to the queue 318 (excellent) of the switch setting control 313.
First) and 312 (normal). Switch setting control 3
13 manages these queues according to their priorities, and sends requests to the MAN switch 1.
0, more specifically, to switch control data concentrator 136.
At normal load, queues 318 and 312 are almost empty and requests are
Always immediately, and usually to the appropriate MAN switch controller.
Therefore, it is processed. In the overloaded state, queues 318 and 312 are
Suppress transmission of low priority packets and maintain relatively fast transmission of priority packets
Means. Experience has shown that if necessary, a request from a normal queue is sent to its destination
If a priority packet for NIM is received, it can be moved to its priority queue.
Noh. Requests placed in queues 318 and 312 include IL, ILH, and circuit
It does not bind the output link of the switch 10. This is the MAN switch controller
Roller 140 (7th This is in contrast to the requests in the queues 150, 152 (FIG. 8) of FIG.   Switch setting control 313 establishes connection within switch 10
If so, it notifies the NIM Queue Manager 311. ILH1
7 is the F in the switch unit queue 314 from the NIM queue manager 311
Data is received from the I / O queue 316 and transmitted to the circuit switch.
Identify the memory locations of the queues of the packet and, for each packet,
Identify a list of one or more ports on which the packet should be transmitted
You. The NIM queue manager then sends a port to ILH17 at the beginning of each packet.
Number and send data from memory 18 to switch 10 for each packet.
Instruct you to The ILH starts transmitting the next queued packet, and this task
Is completed, the connection in the circuit switch is disconnected by the switch setting control 313.
That the data may be transmitted to the memory manager 302. To identify the memory block that can be released.   MINT central control, each with one or more input / output ports
Use multiple fast processors. The specific processes used in this implementation
Is a transputer manufactured by INMOS
. This processor has four input / output ports. This processor is MINT
The processing demand of central control can be satisfied.   Packets come in four XLHs 16. Four XLH managers 305, outgoing
The destination checker 307, the router 309, and the OA & M MINT processor 315
Exists but corresponds to an individual XLH in the MINT. These processors are individually
Operate in parallel to process data entering the XLH of the
Increase the total data processing capacity of Troll.   A header for each packet entering the XLH is sent with the address
But here this The packet contains the header's cyclic redundancy code (this is performed by the XLH).
If the CRC) passes the hardware check, the associated XLH Manager
305 directly. If it does not pass the CRC check,
The XLH is destroyed by the XLH, and the XLH
Reuse the work. XLH manager assigns to headline and its packet
The identified memory is passed to the source checker 307. XLH Manager
, Either a source checker, router, or NIM queue manager
If you find that it is not possible to transmit the packet to its destination,
Recycle recycling. Note the recycled memory block
Used before a memory block allocated by the remanager. Departure
The source checker 307 determines whether the source of the packet is correctly logged in,
And whether the source has access to the virtual network of the packet. Departure
Shingen checker 307 relates to a packet including a packet address in the MINT memory.
Passes the information to router 309, which identifies this packet group.
Into the virtual network name and the destination name of the packet.
It is determined whether a packet should be sent. Router 309 identifies output link
Passed to NIM Queue Manager 311 but by the four XLHs of this MINT
And concatenate packets that are received by a single common output link
Turned to After the first packet to the NIM queue is received, the NIM queue
The manager 311 sends a switch setting request to request a connection to this NIM.
Sent to switch setting control 313. NIM Queue Manager 311 is Sui
Switch these packets in FIFO queue 316 of unit queue 314
When a switch connection is established in the circuit switch 10, all of these packets
Will be sent over this connection at once. Switch control 22
Output control The signal distributor 138 sends an acknowledgment when it has set up the connection. This positive notification
Is received by the switch setting control 313 and the control 313 is
This is notified to the NIM queue manager 311. NIM queue manager next
Connected to ILH17 so that ILH17 can send out all of those packets
Report a list of packets. ILH 17 is the concatenated packet of this set.
When the transmission through the circuit switch is completed, this is the switch setting control 3
13 is notified of this disconnection in the switch 10 and the memory manager 3
01 is used to store the data of this message.
Notifies that a new message has become available. Memory Manager 30
1 sends this release information to the memory distributor 303.
It is distributed to various XLH managers 305 for allocating memory to LH.   The originating checker 307 also operates, supervises and maintains billing information (OA & M).
) MINT process Pass to the server 315, which charges for this packet and MIN
Used to gather appropriate statistics to check the data flow in T
You. This statistic is later combined with other statistics in the MAN network. Router 309
Also, the (OA & M) MINT processor 315 has an OA & M MINT processor.
Can track data about packet destinations for subsequent traffic analysis
To notify the destination of this packet. These four OA & M MINT processors
The output of the server 315 is sent to the MINT OA & M monitor 317, which
Summarizes the data collected by the four OA & M MINT processors and later
, OA & M central control 352 (FIG. 14).   The MINT OA & M monitor 317 also outputs the data of the router 309 to the OA & M.
OA & M Central Control 3 to change via MINT processor 315
Information is received from 52. These changes may require additional terminals added to the network, certain physical
Poe A logical terminal (that is, a terminal associated with a particular user) from one port to another port
Or removal of the physical terminal from the network. Data is OA &
MINT operation, OA & M monitor and OA & M M from M central control 352
The message is sent to the sender checker 307 via the INT processor 315.
Data includes logical user passwords and physical ports, as well as individual logical users.
It contains data on the privileges of The.4.8 MINT operation, management and maintenance control system   FIG. 15 shows a block diagram of the maintenance and control system of the MINT network.
. The operation, management and maintenance (OA & M) system 350 comprises multiple OA & M central controllers.
Connected to roll 352. Each of these OA & M controls has multiple MINs
T, and within each MINT, the MINT central control 2
0 MINT OA & M monitor 317. OA & M system 350?
Many of these messages are all MINTs These various OA & M central controls must be distributed to
Are interconnected by a data ring. This data ring is
Source module identification and therefore the output link of each physical port added to the network
This information is transmitted to the router process of each MINT in the MAN hub.
Stored in the server 309.5. Link 5.1 Link requirements   The link in the MAN system is between EUS and NIM (EUSL) (link 14)
And to transmit data between the NIM and the MAN hub (XL) (link 3)
used. The behavior and characteristics of the data transmitted on these links are specific to each application.
Therefore, although there are some differences, the format used on these links is the same.
You. Having the same format allows for common hardware and software
Can be used.   This link format offers the following features: Designed to be. 1． This provides a high data rate packet channel. 2． It is compatible with the proposed Metrobus "OS-1" format. 3． The interface is simple due to the word-oriented synchronization format.
You. Four. This defines how "packets" should be delimited. Five. This is the CRC for the entire "packet" (and another CRC for the header)
including. 6． This format guarantees the transparency of the data in the "packet"
You. 7． This format is a low bandwidth channel for flow control signaling.
Provide 8． Additional low bandwidth channels can be easily added. 9． Data scrambling provides good transition density for clock recovery
Guarantee.5.2 Description and basis of MAN link   From a performance point of view, the faster the link speed, the better the MNA performance
. The request to make this link as fast as possible is faster with faster links.
Suppressed by the fact that it is expensive. A reasonable compromise between speed and cost is L
ED transmitter (eg, AT & T OLD-200) and multimode fiber
Is to use. The use of ODL-200 transmitters and receivers is
The upper limit is on the order of about 200 Mbit / s. MAN architecture
From now on, the specific data rate of the link may be heavy since the MAN does not perform a synchronous exchange.
It is not necessary. The data rate for the MAN link is based on the Metrobus Lightwave System "O
The data rate is the same as that of the S-1 "link.
Regarding the IEEE International Communication Conference (IEEE Immunational Communication Conferen
ce), 1987, paper 30B. 1.1. S. Schafer (M.S.Schafe
r) (Synchronous Optical Network for Metrobus Lightwave Network).
ptical Transmissi on Network for Metrobus Lightwave Network)]. MA
Another data rate (and format) that can be used within N
Designated by Bell Communication Research Corp.
There is a specification of SONET which is a link layer protocol.5.2.1 Level 1 link format   MAN networks use the low-level link format of Metrobus. This resource
The information on the link is carried by simple frames that are repeated continuously. This frame
The frame consists of 88-16 second words. The first word is the framing sequence and four
Parity bits. In addition to this first word, the other three words
It is Arabic. These overs used for inter-node communication in a metro bus implementation
Headwords are not used by MANs for Metrobus compatibility. This
Due to the language-oriented nature of the protocol, this use is very simple
You. Parallel load Pure 16-bit shift register can be used for transmission, similar to parallel read
Shift registers are available for reception. 146.432 Mbit / s phosphorus
At data rates, a 16-bit word is transmitted or received every 109 nanoseconds
Is done. This approach relies on a lot of link formatting hardware.
It allows realization at the next TTL clock speed. Words of this protocol
The oriented nature of this link, however, places some restrictions on how this link can be used.
I can. To keep the hardware complexity reasonable, the link bandwidth must be 16 bits.
It is necessary to use it in the unit of Gato.5.5.2 Level 2 link format   This link moves the "packet" which is the basic unit of MAN information transmission.
Used for In order to identify the packet, the format is "SYNC" word
And the specification of the word "IDLE". While the packet is not being transmitted, this "IDL
E "word is all the words that make up the basic channel bandwidth (Words not reserved for other purposes). The packet is the leading START_S
Delimited by YNC and trailing END_SYNC word. This scheme is
Works well as long as words with special meaning are not included in the data in the packet.
Restricting the data that can be sent in a packet is an undesirable restriction.
Therefore, transparent data transmission technology must be used. MA
N-link uses a very simple word insertion transparency technique. Packet
Occurrence of a word with special meaning, such as START_SYNC word, in the data
Is preceded by another special word "DLE". This word insertion transformer
Parity was chosen for its ease of implementation. Required for this protocol
The required logic is required for bit insertion protocols such as HDLC
Simpler and slower than. The technology itself is used within the IBM BISYNC link.
Similar to the proven technology used. Used to ensure transparency
Inserted words In addition, if the originating data rate is slightly lower than the link data rate, "FIL
The L "word is inserted.   The last word in any packet is a cyclic redundancy check (CRC) word. This word is
It is used to guarantee this detection if there is data loss in the packet.
This CRC word is inserted into the data stream for transparency or other purposes.
For all of the data in the packet, except for special words such as "DLE"
Is calculated. The polynomial used to calculate the CRC word is the CRC-16 standard
is there.   Before transmission, to ensure good transition density for the optical receiver
All data is scrambled (eg, block 296 in FIG. 13). This
The scrambling of a long sequence of 1s or 0s
Although this is complicated in the data, the probability of transmission on the link is reduced.
Scrambler and descrambler (for example, the block shown in FIG. 12) 252) are well known in the art. The descramble design is self-synchronous
, This is due to occasional bit errors without restarting the descrambler
Allows you to recover.5.2.3 Low speed channel and flow control   All of the payload words in the Level 1 format
Not used for Carry Level 2 format. Additional channels
Included on the link by using specific words exclusively in the frame. these
The low-speed channel 255.295 (FIGS. 12 and 13) controls the MAN network.
Used for the purpose of Packets similar to those used on the underlying data channel
A partition scheme is used on these low-speed channels. Configure low-speed channels
Dedicated words are also used for individual low-bandwidth channels, such as flow control
Is divided into individual bits for each channel. Flow control channel
MAN EUS (between EUS and NIM) Used to provide hardware-level flow control over L
You. This flow control channel (bit) from NIM to EUS is E
Indicate to the US link transmitter whether this is allowed to carry further information
. NIM indicates that the EUS transmitter has been flow-controlled and is actually shut down
Is designed with enough memory to absorb all of the data transmitted by Day
Data transmission is stopped between packets or in the middle of packet transmission. Packet
Between packets, the next packet is transmitted until flow control is released.
Absent. If flow control is performed in the middle of a packet, the data transmission
Stop sending immediately and start sending “Special FILL” codeword.
It is necessary to start. This codeword, like other words, is
It appears in the body and is escaped by the "DLE" code word.6. System clocking   MAN switches are described in section 3 Asynchronous spatial switch with very fast setup controller
Tissue (asynchronous space switch fabric). Data organization for this switch
(data fabric) is digital from DC to data rate over 200 Mbit / s
It is designed to propagate signals with high reliability. Many through this organization
The total bandwidth requirement of the MAN hub is set by this organization because the routes can exist at the same time.
Easy to fill. However, this simple data organization is completely flawless
Do not mean. Due to the mechanical and electrical constraints to achieve this organization,
Not all paths through a given switch will receive the same amount of delay
No. Variations in the path delay between the various paths are the bit times of the data passing through all of them.
Is so large that synchronous exchange is not possible. Features in MINT
At any time when a path from a fixed ILH to the output port of the switch is established
Data transmitted through this path propagates on the previous path through the switch.
Same as data sent There is no guarantee that they have one relative phase. To use this high bandwidth switch,
Therefore, the data coming out of the switch port is used for the synchronization link to the NIM.
It is necessary to synchronize very fast to the clock used.6.1 Phase matching and scrambler circuit (PASC)   Synchronize the data coming out of the switch and drive outgoing link to NIM.
The unit to be inverted is a phase alignment and scramble circuit.
mbler Circuit, PASC) (block 290, FIG. 13). ILH
And the PASC circuit have the same master because the MANs are all part of the MAN hub.
It is possible to distribute the clock to all of these. This has some advantages
One. The same clock base used for transmitting data from the ILH in the PASC
By using the standard, the data is transmitted to the PASC, which
You can be assured that you will not enter at a faster rate than is done. This is a large Eliminates the need for smart FIFOs and sophisticated elastic memory controllers. PAS
The fact that the bit rate of all data entering C is exactly the same
Make it easier.   ILH and PASC are decentralized resources for the format described in the previous section.
It can be thought of as a distributed link handler. I
LH generates a basic framing pattern into which the data is inserted,
Send to PASC through this organization. PASC uses this framing pattern
Matches own framing pattern and merges into low-speed control channel
And then scramble the data for transmission   The PASC inserts the incoming data by adding an appropriate amount of delay into the data path.
To synchronize with the reference clock. In order for this to be successful, ILH must
The reference clock slightly advanced from the reference clock used by the PASC.
Must be sent by lock. ILH requires The number of advanced bit times is the actual maximum received during transmission from the ILH to the PASC.
Determined by small delay. Delay that PASC can insert into the data path
Amount of path delay can occur for different paths through the switch
Depends on the fluctuation.   FIG. 23 is a block diagram of an exemplary embodiment of the present invention. Inconsistent data
Data enters the tapped delay line 1001. The various taps on the delay line
REFCLK which is 180 degrees out of phase with respect to the reference clock (REFCLK)
The signals named edge sampling latches 1003,. . . , 100
5 clocked. The output of this edge sampling latch is
The output of the unit 1007 is supplied to the
Used to control 1013. Select logic 1007 includes latches 1003,. .
. , 1005 includes a set of internal latches for repeating the states. The selection logic is
Highest rank order entry carrying logic "1" , And a priority circuit connected to these internal latches. This output is
Is the coded identification of the connected input of. This selection logic 1007
Signal, ie, one release signal and the internal latch of the group of selection logic.
With one signal from all. Release signal is zero between two data streams
Then, the internal latch receives a new input. Responds to the first pulse of the data stream
In response, the edge sampling latches 1003,. . . , 1005 first "
After the input of "1" is received, this transformer is used until the release signal returns to zero.
The state of the parent latch is maintained. This release signal recognizes the presence of a data stream
Set by the out-of-band circuit.   The output of the tapped delay line also includes a series of data latches 1009,. . . 1
Enter 011. The input to this data latch is clocked by the reference clock.
Is done. Data latches 1009,. . . , 1011 to the selector circuit 1013
Entering The output of one of these data latches from selection logic 1007
Select based on input and connect this output to the output of selector 1013.
Is a bit aligned data stream as named in FIG.   Once the bits are matched, they are provided to driver XL3 along with the tapped output.
It is fed into a shift register (not shown) for feeding. This means that the data stream is 1
This is to enable transmission synchronously starting at a 6-bit boundary. Send this digit
The operation of the register and auxiliary circuit is substantially the same as the operation of the tapped delay line configuration.
Are identical.   The selection logic is implemented by a commercially available priority selection circuit. The selector is
It is simply an 8-choice selector controlled by the output of the selection logic. 16 choices
If you need a fine matching circuit to use, this can be easily done using the same principles.
realizable. The configuration described here has one common source clock,
This is particularly attractive when the length of the data stream is limited. Both The source clock is required because the clock is not derived from the incoming signal,
Used to properly gate incoming signals. The clock length limit is
If a particular gate selection is kept for the whole block and the block is too long,
Some phase will cause loss of synchronization and bits will be dropped
Required.   In this embodiment, the signal is passed through a tapped delay line and
Sampled by lock and inverted clock, but tapped clock
Method of sampling through a delay line and using a delayed clock
Can also be used by the application.6.2 Clock distribution   The operation of the MAN hub is for all ILH and PASC units in the system.
Very dependent on the use of a single master reference clock. The master clock is
Must be correctly distributed to units. This to be distributed
Basic In addition to the clock frequency, the frame start pulse is distributed to the PASC and
The start-of-frame pulse must be distributed to the ILH. All these machines
A single clock distribution link (fibre or twisted
Line).   The information carried on these clock distribution links comes from a single clock source.
You. This information can be split in the electrical and / or optical domain and transmitted to as many destinations as needed.
Wear. Keeping the phase of the information on all clock distributors perfect is the ILH and P
ASC was attempted because of its ability to correct these phase differences regardless of the cause.
Absent. The transmitted information is simply alternating 1s and 0s with two exceptions. In line
The occurrence of two 1s indicates an advance frame pulse and the occurrence of two 0s in the row
Indicates a normal frame pulse. One of these clock distribution links terminates
Each board contains a clock recovery module. This clock recovery module
Same as the module used for the link itself One. The clock recovery module provides a very stable bit clock
The additional logic, on the other hand, is
Extract the frame. The clock recovery module has the correct frequency without bit transitions.
Oscillation continues for several bit times in number, so bits with very low probability of occurrence
  Even errors do not affect the clock frequency. This frame or ad
The logic to search for a balanced frame signal also knows that the frame pulse is periodic.
And the extraneous pulses caused by bit errors are negligible.
Can be designed to withstand lars.7. Web interface module 7.1 Overview   One network interface module (NIM)
Or multiple end user system links (BU)
SL) to one MAN external link (external kink, XL). Like this
To By doing so, the NIM is able to handle network transaction units (ie, packets and
SUWU), and performs individual output packets.
By adding a physical “transmission port number” to the port, the integrity of the source identification is ensured.
I do. The latter function is combined with the network registration service described in section 2.4,
In order to gain access to network-provided services for which the user has no rights,
Prevent impersonation. NIM thus gives the boundaries of the MAN network body,
Owned by the network provider, while the UIM (described in Section 8)
Owned by the itself.   In this section, the basic functions of NIM are described in more detail,
The M architecture is shown.7.2 Basic functions   The NIM must perform the following basic functions:EUS link interface : One or more interfaces are EUS
Link (section 2.2.5). Downstream link (ie NIM to UI
M link) is the upstream link flow when the NIM input buffer is full
  The data channels and channels used by the NIM to control
Out of band channels. The downstream link is flow controlled
Therefore, the upstream link flow control channel is unused. De
Data and header check sequences (DCS, HCS) are sent by the UIM on the upstream link.
And checked by the UIM on the downstream link.External link interface : XL (Section 2.2.6) is very similar to EUSL
, But without DCS checking and generation at both ends. This is d
Errors, but allows for delivery of potentially valid data to the UIM
I do. The destination port number in the network transaction unit arriving on the downstream XL is
The data is discarded if checked by NIM and found to be invalid. It is.Concentration and demultiplexing : Network transaction arriving on EUSL
The knits compete and are statistically multiplexed to the outgoing XL. Net transa arriving on XL
The action unit maps the destination port number to one or more EUS links.
By pinging, it is routed to the corresponding EUSL.Outgoing port identification : Each network transaction where the port number of the originating UIM goes upstream
The port number is added to the head of the unit by a port number generator 403 (FIG. 16).
This port number is provided by MINT, for the "fraudster" service (the most basic
(Including data transmission services) to prevent unauthorized access to
Checked against7.3. NIM architecture and operation   The architecture of the NIM is shown in FIG. In the following subsections
The operation of the NIM will be briefly described. 7.3.1 Upstream operation   The ingress transaction unit sends these EUSL interfaces from the UIM.
Received at the receiver 402 of the source 400 and the serial / parallel converter 4
The words are converted into words in the buffer 04 and stored in the FIFO buffer 94. Individual EUSL
The interface is connected to a NIM transmit bus 95, which is a parallel data bus.
Data arbitration and various signals for bus arbitration and clocking. Net tiger
If the transaction unit is buffered, the EUSL interface 400 sends
Arbitrate access to bus 95. This arbitration runs in parallel with the data transmission on the bus.
Be done. When the current data transmission is completed, the bus arbiter competes.
The right to use the bus is given to one of the corresponding EUSL interfaces. Individual transa
To the beginning of each packet by the port number generator 403
The entered EUSL port number is transmitted first, followed by the network transaction.
Unit is transmitted. XL interface Within 440, XL transmitter 96 provides the bus clock and switches from parallel to serial.
Perform conversion 442 to real and transmit data on upstream XL3.7.3.2 Downstream operation   The network transaction unit from MINT arriving on downstream XL3 is XL
Received by XL receiver 446 within interface 440,
Connects to NIM receive bus 430 via serial / parallel converter 448.
Continued. The receive bus is similar to, but independent of, the transmit bus. Also,
The EASL interface is connected to the reception bus via the parallel / serial converter 408.
A base transmitter 410 is connected. XL receiver performs serial / parallel conversion
It provides a receive bus clock and also provides incoming data on the bus. Individual
The EUSL interface decodes the EUSL port number associated with the data and requires
If so, transfer this data to EUSL. If necessary, use multiple EUSL
Interface Data can be transmitted as a broadcast or multiplex broadcast operation. Individual decoder 40
9 checks the reception bus 430 when the port number is being transmitted, and
Is intended for the end user of this EUSL interface 400.
And if so, this packet is delivered to the transmitter 410 to EUSL14.
Transferred to reach. Invalid port number (for example, error code ski
Violation of the system) results in the destruction of the data (ie, the EUSL interface).
Not transferred by the face). The decryption block 409 is for a specific EUS link
The directed information is transmitted from transmit bus 95 to parallel / serial converter 408 and
Used to gate to transmitter 410.8. Interface to MAN 8.1 Overview   The user interface module (UIM) is one or more end-user
User system (EUS), local area network (LAM), or dedicated
A simple point-to-point link C for connecting to one MAN end user system link (EUSL) 14
It consists of hardware and software. Throughout this section, the term E
US is used to generically refer to these network end-user systems. clear
In addition, the part of the UIM used to connect a particular type of EUS to the MAN is:
The architecture of this EUS, as well as the required performance, flexibility,
And the cost of implementation. However, some of the functions provided by UIM
Must be provided by all UIMs in the system. Therefore, U
The architecture of IM has two different parts:
The network interface to provide and the rest for the particular type of EUS connected
It is convenient to think that it consists of an EUS interface that realizes the UIM function of
.   Not all EUSs require specific performance on dedicated external links. NI
Concentrations provided by M (described in Section 7) are severe. Response time requirements as well as those required to efficiently utilize full MAN data rates.
Traffic with instantaneous I / O bandwidth and required to load XL efficiently
Is a suitable way to provide access to multiple EUSs that do not generate As well
Some EUS or LAN with some intermediate links (or LAM itself)
To the same UIM. In this scenario, the UI
M functions as a multiplexer, and a plurality of M parallel to one network interface
Provides an EUS (actually LAN or link) interface. This way
Do not allow direct connection to these system buses, and
Suitable for EUS providing only one link connection with width. End users
These multiplexes or concentrators are provided at the UIM, and the MAN has more multiplexing.
Or provide subscriptions at NIM.   This section covers the UIM network interface and the EUS interface.
Both ar Describe the architecture. Describes the functions provided by the network interface
And its architecture is shown. Heterogeneous mixture of EUS connected to MAN
Due to the nature, general treatment of the EUS interface is not allowed. Instead, EUS
Interface design options are shown and one possible EUS interface configuration
One specific example, EUS, is used to illustrate the scale.8.2 UIM-network interface   The UIM network interface implements functions independent of EIM of UIM. Individual
One or more EUS interfaces can be connected to a single MAN
Connect to EUSL.8.2.1 Basic functions   The UIM network interface must perform the following functions.EUS link interface : Optical transmission is used for the interface to the EUS link.
Transceivers and receivers and link level functions required by EUSL (eg
For example, CRC generation and checking, Data formatting, etc.).Data shock : Outgoing network transaction unit (that is, packet and SUW
U) buffers these so that they can be transmitted on the high speed network link without gaps
Is needed. Incoming transaction units provide speed control and level
Buffered for protocol 3 (and more) protocol processing.Managing buffer memory : One LUWU packet is another at the receiving UIM
May arrive between LUWU packets. Some LUWU's this
To support simultaneous reception, the network interface uses this receive buffer memory.
And dynamically connect incoming packets as soon as they arrive to LUWU.
I need to be able to do it.Protocol processing : Outgoing LUWU is not fragmented into packets for transmission in the network.
I have to. Similarly, incoming packets are received by EUS Must be rejoined to the LUWU for delivery to.8.8.2 Architectural options   Obviously, all of the features listed in the above subsection are optional EUS
Must be performed to interface to MANEUSL. However
And where these functions should be performed, that is, they are performed within the host.
Architectural choices as to what should be done or done externally
Is required.   The first two functions, for different reasons, can be located outside the host.
Required. Introduce the first lowest level feature, the MAN EUS link
Interface simply consists of dedicated hardware that is not part of the general EUS
Must be implemented outside the host for that reason. EUS Link Inn
Interface is simply a bidirectional I / O port to the rest of the UIM network interface.
Function. On the other hand, the second function Data buffering can be implemented in existing host memory due to overwhelming bandwidth requirements.
I can't do that. On reception, the network interface will receive incoming packets or
SUWU buffering at full network data rate (150 Mbit / s)
Required. This data rate typically places incoming packets directly into EUS memory.
It is impossible speed. Similar bandwidth constraints apply to packet and SUWU transmissions.
This also applies to feeds, where they are completely buffered and then at full speed 150
This is because it must be transmitted at M bits / sec. These restrictions are
Request to provide the necessary buffer memory on the outside. FIFO is for transmission
Is sufficient to provide the speed adjustment required for
ー Due to lack of control and reception packets being interposed,
Providing large amounts of random access memory as buffer memory
Is required. For MAN, the size of the receive buffer memory is 256 keys.
Lobite to IM A range of bytes is possible. The specific size is determined by the host's interrupt wait time and host
G depends on the maximum size LUWU allowed by the software.   The last two functions are conceptually processes that can be performed by the host processor itself.
With reason. The third function, buffer memory management, is the
Includes timely allocation and deallocation of resources. Waits associated with assignment behavior
Time requirements are very stringent, but again this is where high data rates and packet folding
Due to the possibility of arriving in return. However, this is a few blocks of memory in advance.
(For a suitable burst size). Follow
It is possible for the host processor to manage the receive packet buffer
. Similarly, the host processor executes the fourth function, MAN protocol processing.
You can carry it or not.   The location of these last two functions determines the level at which the EUS connects to the UIM
. The host CPU manages the packet buffer memory and If you assume responsibility for the protocol processing ("local" configuration), the EUS interface
The unit of data transmitted across the interface is a packet, and the host
Responsible for WU fragmentation and recombination. On the other hand, these functions are
To the processor of the (frontend processor, FEP
) Configuration), the unit of data transmitted across the EUS interface is LUWU
It is. Theoretically, under the interleaving constraint at the EUS interface
, The unit of data transmitted can be any amount less than or equal to the entire LUWU
Units delivered by the transmitter and accepted by the receiver
It does not need to be the same size as
As a uniform solution, LUWU may be a basic unit. The FEP configuration is
Releases most of the processing responsibilities from the host CPU and provides a high level EUS interface.
Providing networking information hides the details of network operation from the host. FEP provided
If the host Knowing only about LUWU, these transmissions and receptions are more CPU intensive.
Can be controlled at no level.   It is also possible to implement a low-cost interface using local configuration.
Network interface architecture as described in the following sections
Is required by the high performance EUS, an ordinary user of the MAN network.
This is an FEP configuration. Another reason to choose the FEP configuration is that
Suitable for interfacing an AN to a LAN, eg ETHERNET
But in the latter case, provide buffer memory management and protocol processing
"Host CPU" does not exist.8.2.3 Network interface architecture   The architecture of the UIM network interface is shown in FIG. The following
By indicating scenarios for sending and receiving data,
Simple operation of UIM network interface Just explain. Here, an FEP type architecture is adopted. Toes
Management of receive buffer memory and processing of MAN network layer protocol
Is performed outside the EUS host CPU.8.2.3.1 Data transmission   The primary responsibility of the network interface in transmission is to send user work of any size.
Unit (user work uint, USA) is fragmented into packets (if necessary)
And encapsulate the user data within the MAN heading or trailing edge.
Data to the network. Request LUWU transmission to start transmission
Is sent to the EUS interface, and the network interface
The processing unit 450 is processed by the memory processing unit 405.
It also implements a protocol processing function. Interface processing for individual packets
The protocol processing part of the device 450 creates a heading and sends it to the transmission FIFO 15.
Write in. Data for this packet Is next transmitted through the EUS interface 451 and in the link handler 460.
It is sent into the IFO 15. When this packet is completely buffered, link hangs
Dollar 460 sends this out on the MAN EUS link using transmitter 545.
Subsequently, the trailing edge calculated by the link hydrator 460 is sent out. Re
Link does not cause NIM packet buffer overflow
Flow control. This transmission process is performed for each packet
Repeated. The transmission FIFO 15 is used to transmit packets at the maximum speed.
Has space for two maximum length packets. When the user completes the transmission,
The notification is received via the EUS interface 451.8.2.3.2 Receiving data   150 Mbit / s link rate received by incoming data receiver 458
Is loaded into the elastic buffer 462. Dual Port Video RAM
For the communication buffer memory 90 Data is used from this elastic buffer to the shift register in the receive buffer memory.
464 via its serial access port. Individual package
The next bit from this shift register is the main memory array 4 of the receive buffer memory.
At 66, it is transmitted under the control of the receiver DMA sequencer 42. Perform this transfer
The block address used to perform the
Provided by the processing unit 450 via the buffer memory controller 456.
It is. The buffer memory controller 456 may be imposed by the folded SUWU.
Fewer addresses in hardware to mitigate demanding latency requirements
Buffer. Block 450 comprises blocks 530, 540, 5 shown in FIG.
42, 550, 552, 554, 556, 558, 560 and 562.
The network interface processor stores the random access port in the buffer memory.
Headings are not stripped because they have direct access through the
With buffer Put in memory. The reception queue manager 558 in the processing device 540
Handle the process and use the input from the memory manager 550 to arrive
It tracks SUWU and LUWU. Use EUS to check the arrival of data
Notified by the face processing device 450 via the EUS interface. E
Details of how data is distributed to the US can be found in the specific EUS
Interface, and is described in Section 8.3.3.2 as an example.
You.8.3. UIM-EUS interface 8.3.1 Thought   This section describes half of the EUS-dependent network interfaces.
You. The basic function of the EUS interface is the EUS memory UIM network interface.
To transmit data in both directions between Each specific EUS interface
The protocol that performs the transmission, the data and control message
It defines the format and the physical path to controls and data. The individual sides of this interface are designed to protect themselves from overruns.
A low control mechanism must be implemented. EUS is its own
The network must be able to control the flow of data to and from the
You are required to protect your body. At this basic functional level, the EUS interface
Only the commonality of the faces can be stated. EUS interface
, Due to the variety of EUS hardware and system software,
Not. The relationship between the demand for applications that use the network and the ability of EUS
The choice of interface design is required in the context of performance and flexibility.
Various interface choices exist for only one type of EUS
I do.   The choice of this set is simple if the interface hardware has few elements
From complex designs to complex designs with complex buffer and memory management schemes
Means spanning. The control functions in the interface are simple E
US Interface to network level 3 protocol and even distributed apps
Interface to handle even higher level protocols for application
Ace. The software in EUS is also the existing network software.
From simple data transmission schemes that fall under the network to very flexible use of the network
Or more complex new features that enable the best performance that the network can provide.
There is even a choice up to EUS software. These interfaces are specific existing
Need to be designed for EUS hardware and software systems
Required for network applications running within these EUSs.
It is also necessary to analyze the relationship between the convenience provided by
You.8.3.2 EUS interface design options   The trade-off between preprocessor (FEP) and EUS processing is the same basic function
Is an example of a different interface approach to achieve Consider the diversity of receive buffers. Dedicated with high performance system bus
Network packet messages directly from the network link in the EUS architecture
You can also receive it. However, this interface is usually
Before delivering the message to EUS memory, the packet message
Perform buffering. Usually, EUS transmitting to or receiving from the network
, Do not know (or do not want to know) about internal packet messages
. In this case, the receiving interface is the normal size between the sending EUS and the receiving EUS.
It is necessary to buffer a plurality of packets from the LUWU of data as a transmission unit.
Required. Each of these three reception buffer situations is possible, and different data can be stored in the EU.
Requires a significantly different EUS interface for transmission to S-memory. EU
S has a specific need to process network packet messages and this task
If you have the processing power and system bus performance that can be dedicated to
Interface EU The S-dependent part is simple. However, this processing is usually applied to the EUS interface.
To improve the EUS performance.   Various transmission buffering approaches also provide trade-offs between FEP and EUS processing.
Clarify the problem. In specialized applications, high-performance processes
EUS with a processor and bus sends network packet messages directly to the network
Can be. However, this application is more than packet message size
If you always use a long EUS transaction size, this is the packet itself.
The EUS process can be too burdensome to generate the message. FE
P to help format this level 3 network protocol.
You can also receive. This is because EUS is freed from internal network message size.
Or if this is a different set of network applications with very different transmission sizes.
This is true even if you have applications.   EUS hardware architecture and Depending on the level of performance required, data between the EUS memory and the network interface
To transmit data, the choice between programmed I / O and DMA is determined. Step
In the programmed I / O approach, perhaps the control signal and data
Are transmitted on the same physical path. In the DMA approach, EUS
Some sort of transmission of control information in the interface protocol
A shared interface is used, and the EUS interface
Buffered through EUS system bus without using the Sessa cycle
Used by DMA controller to transfer data between memory and EUS memory
Is done.   There are several alternatives for the location of the EUS buffer for network data. data
Is on a preprocessor network controller circuit board with its own private memory
Can also be buffered. This memory uses a bus that uses DMA transmission for EUS.
To connect to EUS and access via bus Connected to a dual port memory
Connected to the dual port memory located on the CPU side of the bus that uses
Can also be. This application now needs access to the data.
A variety of technologies can be used, but some technologies require the end user's workspace.
Address space used by UIM to store this data directly
Is mapped to Some technologies also require that the operating system
Buffer the data and copy it back to the user's private address space.
Also require   EUS responsible for transmitting control and data information over the interface
Several options also exist for designing driver-level software within
. This driver also implements EUS interface protocol processing.
Or just have the bits transmitted over the interface. Do
In order for the driver to run efficiently, protocol processing in the driver must be performed. Sometimes it is too flexible. Flexible based on specific application
EUS interface protocol processing to increase the
You can move to a higher level. The closer you get to the application, the more
More intelligence is given to interface decisions at the expense of EUS processing time
It is. EUS to transmit data to or receive data from the network
Various interface protocol approaches, for example, priority
, Or privileges. You need this flexibility
Network applications that do not have a more direct interface to drivers and the network.
Source can be used.   As can be seen from above, hardware and hardware at various levels in the system
Various choices are allowed in both software and software.8.3.3 Implementation example: SUN workstation interface   Explains the EUS-dependent part of the interface For this purpose, one particular interface is described here. This interface
Is manufactured by Sun Microsystems, Inc.
A workstation based on the Sun-3 VMEbus. This is a single
Are connected to a single network interface. This EUS is also
Allows direct connection to this system bus. UIM hardware is V
The ME bus system is considered to be a single circuit board plugged into the bus.   First, the Sun I / O architecture is described, and then the interface
Hardware, interface protocols, and new and existing
Describes the choices in designing a connection to network application software.
It is.8.3.3.1 Sun Workstation I / O Architecture   Sun-3 I / O architecture based on VMEbus structure and memory tube
Management unit (memory management uint, MMU) D called direct virtual memory access (DVMA)
Provides MA approach. FIG. 17 shows Sun DVMA. DVMA is a system
Device on the system bus directly DMAs to Sun processor memory
And the main bus master can
  Allow bus slaves to DMA directly. This is on the system bus
The address used by the device to communicate with the shell is similar to that used by the CPU
This is called a “virtual” because it is a virtual address. This is the DVMA approach
All addresses used by devices on a given bus are
Are treated as if they were virtual addresses generated by the CPU.
Also ensure that The dependent decoder 512 (FIG. 18) provides the VME bus address
Lowest megabyte of space (0x in 32-bit VME address space)
0000 0000 → 0x000f ffff), this megabyte is
System virtual ad Most significant megabytes of address space (0 in 28-bit virtual address space)
xff0 0000 → 0xffff ffff). (0X follows
Indicates that the character is an octal character. ) The driver debugs the buffer address.
When it needs to be sent to a device, this means that the address that the device places on the bus is V
28 bits to fit within the low megabyte (20 bits) of the ME address space
The higher 8 bits must be stripped from the address.   In FIG. 18, a CPU 500 drives a memory management unit 502.
. Then, the memory management unit 502 controls the VME bus 504 and the buffer 508.
The on-board memory management unit 502 includes a VME bus 504 and a buffer.
508 is connected to the on-board memory 506. VME bus is a DMA device
Communicates with chair 510. Other on-board bus masters, such as ETHE
The RNET access chip also accesses the memory 508 via the MMU 502
It is possible You. Thus, the device will have DVMA space in these low (physical) memory areas.
DVMA transmission can be performed only in the reserved memory buffer.
However, the shell does not provide redundant mapping of multiple virtual addresses to physical memory pages.
to support. Thus, a page in user memory (or shell memory) is
(Or this) should appear in the address space of the process requesting the operation.
(As it comes from) can be mapped into the DVMA space. This
Liver directly creates a shell page map that supports user space DVMA.
Use a routine called mbsetup to set up.8.3.3.2 Sun UIM-EUS Interface Approach   As mentioned above, there are many options for designing interface properties.
Exists. The DMA transmission approach was designed using the Sun-3 interface.
Was. Interface with FEP capability, high performance matching system bus
Inn Interface and various new and existing network applications use the network.
The flexibility and the like of the EUS software that enables the EUS software have been described.   Sun-3 is a window system, and has been upgraded to support multiple users.
A system with potentially many simultaneous processes running. DMA and FEP
The approach reduces the burden on the Sun Processor while network transmissions are taking place.
Selected for UIM hardware is plugged into the VMEbus system bus
Can be considered as a single circuit board. Connects directly to the system bus
Are required to try to be the most powerful interface possible.
It is. Sun's DVMA efficiently transfers data to and from processor memory
To provide a means for moving to. UIM (Fig. 4) contains UIM
Transfers data to memory and transfers data from EUS memory to UIM via bus
There is a DMA controller 95 for transmission and a host A shared memory interface for transmitting control information in the interface protocol
Interface is also conceivable. The Preprocessor (FEP) approach uses data from the network.
Data is transmitted to the EUS at a higher level. Level 3 protocol
Processing is performed and the packet is converted to a medium sized unit for user transmission.
LUWU. A variety of network applications running on the Sun
Therefore, the FEP approach is based on EUS software
It does not need to be tightly bound.   This Sun-3 DVMA architecture allows for EUS transaction size
Limit to a maximum of 1 megabyte. If the user buffer is not locked in
Alternatively, the shell buffer can be used as an intermediate step between the device and the user.
However, in this case, performance is sacrificed for the copy operation. “Mbsetup
If the transmission is made directly to user space using the “approach,
Space is locked in memory and the transmission Throughout the process, this becomes unusable due to swapping. this is,
One tradeoff. This means that it ties up the resources in the machine, but only
However, it is more efficient if copying from other buffers in the shell can be avoided.
You.   Sun System is an existing network application that runs on ETHERNET,
For example, these Network File System (NFS)
One. Run these existing applications on MAN, and extend MAN
To open up the potential for new applications that use the capabilities provided,
Flexible EUS that can handle various network applications simultaneously
Software and flexible interface protocols are required.   Figure 19 shows the operation and interface functions between MIN, UIM and EUS
FIG. The EUS shown in this particular embodiment is a Sun-3 workstation.
It is. However, these principles are not as simple or complex as other end users.
The Also applies to systems. First, EUS from MINT via NIM and UIM
Think about the direction toward. As shown in FIG. 4, link from MINT11
3 is distributed via a link 14 to one of a plurality of UIMs
The UIM is stored in the reception buffer memory 90 of the UIM, and the data is stored here.
Bus 93 with DMA interface in pipelined form
Is transmitted to the corresponding EUS via. The components required to achieve this data transmission
The trawl structure is shown in FIG. In other words, the input from MINT is MINT
Controlled by the link handler 520 to the NIM. Link handler 5
20 outputs this output from multiple NIMs to UIMs under the control of router 522.
Send to one of the link handlers (N / ULH) 524. MINT / NIM phosphorus
Handler (M / N LH) 520 uses different types of Metrobus physical layer protocol.
to support. The link handler 524 from NIM to UIM is also in this realization.
Metro bus Supports layered protocol, but can support other protocols
Noh. Different protocols may coexist on the same NIM. N / U
  The output of LH 524 is sent over link 14 to UIM 13 where it is N
The IM / UIM link handler 552 relaxes the data in the reception buffer memory 90.
Be hit. The buffer address is supplied by the memory manager,
It manages a list of free and allocated packet buffers. packet
The status of the receipt is obtained by N / ULH 552, which is
The checksum is calculated and verified through the data
Output to the driver 556. The receive packet handler 556 stores this status information in memory.
  Pair this with the buffer address received from Manager 550 and
Put it on the received packet list. The information about the received packet is
Sent to queue manager 558. Receive Queue Manager 558 is packet information
LUWU and SU Assembled in queue for each WU and EUS still informed about it
Hold the queues of LUWUs and SUWUs that are not present. Receive Queue Manager 558
Is EUS / UIM link from EUS for information on LUWU and SUWU
  An inquiry is received via the handler (E / ULH) 540, and
, Notification message via UIM / EUS link handler (U / ELH) 562
Send a page. SUWU is included in the message notifying EUS of receipt of SUWU.
However, this notification completes the receiving process. However
, LUWU, the EUS allocates its memory for reception and issues a reception request.
Issued to the reception request handler 560 via the E / UL LH 540. Reception required
The request handler 560 creates a received worklist and sends it to the resource manager 554.
send. The resource manager 554 controls the hardware and controls the EUS bus 92 (FIG. 4).
) Perform the data transmission performed above via the DMA device. Receiving from EUS
Be sure to request Note that there is also no need for the total amount of data in the LUWU. As a matter of fact
When the EUS makes its first receive request, the data is still at the UIM.
Not all of them have arrived. When subsequent data arrives for this LUWU,
The EUS is notified again and has the opportunity to make an additional receive request. This method
Therefore, data reception should be pipelined as much as possible to reduce latency.
It is. Following transmission of the data, receive request handler 560 switches U / E LH 562.
Notify the EUS via E-mail and deliver the LUWU to the memory manager 550.
Command to deallocate memory for the part. Thus, this
Mori is ready to use new data.   In the opposite direction, that is, in the direction from EUS26 to MINT11,
Is controlled as follows. EUS26 driver 570 sends transmission request
Send to handler 542 via U / ELH 562. In the case of SUWU, this transmission
request It itself contains the data to be transmitted and the send request handler 542 sends this data.
And send it to the resource manager 554. Resource manager 554
Calculate the packet header and store both header and data in buffer 15 (FIG. 4).
From here, this is the flow control program that takes effect on link 14
The UIM / NIM link handover when
It is transmitted to NIM2 by Dora 546. This packet is sent to the UI at NIM2.
Received by M / NIM link handler 530 and stored in buffer 94
It is. Arbiter 532 then NIM / MINT link on MINT link 3
The packet to be transmitted next to the MINT 11 under the control of the handler 534 is
Alternatively, a plurality of buffers 94 in the NIM 2 are selected to select SUWU.
U. In the case of LUWU, the transmission request handler 542 breaks this request into packets.
, Send the send worklist to the resource manager 554. The resource manager 554
For each packet To create a heading, write this heading in the buffer 15, and control the hardware.
To transmit packet data via the EUS bus 92 via DMA.
And instruct U / N LH 546 to transmit the packet when permitted.
I do. The transmission process then proceeds as in SUWU. In either case
, The transmission request handler 542 sends the SUWU or LUWU from the resource manager 554.
When the transmission is completed, the driver 570 receives the notification when the U / E
  Notified via LH 562 and if necessary this transmission buffer is freed
.   FIG. 19 also shows details of the internal software structure of EUS26. Two ties
Of the blocks 572, 574, 576, 578, 580
In one, the user system performs level 3 and higher functions.
Figure 19 shows the U.S. Defense Agency's Advanced Research Project Headquarters network (
Network of the Advanced Research Projects Administration of the U.S. (Department of Defense, ARPAnet)
This includes the inter-net protocol 580 (level 3) and the transmission control protocol.
Call (TCP) and User Datagram Protocol (UDP) blocks
578 (TCP is used for connection-oriented services and UD
P is designed for connectionless service). Higher
At a lower level, a remote procedure call (block 576), a network file server (
Block 574) and a user program 572. Alternatively,
The service of the MAN network is indicated by a blank block 584 between the user and the driver.
User directly interfacing with driver 570 (block 582)
) Can also be called directly by a program.8.3.3.3 EUS interface function   The main functional part of the transmission EUS interface is the control in
Interface, and data between the EUS and UIM through the system bus. DMA interface for transmitting data. When transmitting to a network,
Information and data describing the LUWU or SUWU to be transmitted reside
Information about the EUS buffer is received. This control information from EUS
Contains the address of the destination MAN, the destination group (virtual network), the LUWU length, and the service.
Includes type columns for screw types and high-level protocol types.
The DMA interface sends user data from the EUS buffer to the UIM. this
The network interface formats LUWU and SUWU into packets.
, On the link to the network. This control
Interfaces include multiple pending requests for flow control, priority and preemption, etc.
Can have various variations of UIM is this EUS memory
The amount of data taken from the network and the amount sent to the network are controlled.   On the receiving side, the EUS polls for information about the received packet.
,Control The interface responds with the LUWU information from the packet header and the EUS
Sends current information about how much of the transaction has been reached. Con
Through the Troll interface, the EUS receives data from these memories.
DMA interface sends data from the memory on the UIM to the EU.
Send to S memory buffer. This in the receiving interface protocol
Polling and response mechanisms are available for many EUS
Gives flexibility. EUS is all of the transactions coming from outgoing EUS
Can be received, or a part can be received. This also means that E
Provides a flow control mechanism for the US. EUS is this
Control when, and in what order.
To8.3.3.4 Sun Software   This section describes a typical end-user system, the Sun-3 Workstation.
How A description will be given as to whether or not the MAN is connected. Depends on different end-user systems
Software may be used. Interface to MAN is ratio
It is relatively simple and efficient for many systems tested.8.3.3.1.4.1 Existing network software University of California at Berkely, California
Is derived from the 4.2BSD UNIX system developed by
. As with 4.2 BSD, this is part of the shell, as in the ARPAnet protocol.
Realization, that is, the protocol (IP) between the nets, the connection above the IP
  Transmission Control Protocol (TCP) for oriented services, and
User Datagram Pro for connectionless services on top of IP
Protocol (UDP). The current Sun system puts IP above the network layer
Used as an inter-net sublayer in half. Network layer lower half The segment is the network-specific sublayer. This is currently a specific net hard
Driver-level software that interfaces to the hardware connection,
It consists of an ETHERNET controller, where the link layer MAC protocol is
Is achieved. To connect Sun Workstation to the MAN network,
Must be compatible with existing networking software frameworks
You. Software for Sun's MAN network interface is at the driver level.
  It may be software.   The MNA network is, of course, a connectionless or datagram type network.
LUWU data and control information traverse this interface towards the network
Form an EUS transaction. Existing network services are MAN network datagrams
It can be provided using LUWU as a base. Sun Software
Configures both connectionless and connection-oriented support.
MAN Datagram Network Build application services on top of the service layer. Sun already has a variety of web apps
Application software, the MAN driver multiplexes multiple upper layers
Basic services with flexible flexibility can be provided. This multiplex
The ability to make MAN power more straightforward, not just for existing applications
This is necessary for new applications that will be used indirectly.   Address translation into EUS at driver level in host software
Service function is required. This translates the IP address into a MAN address
It is possible. This address translation service works with the current Sun address
Similar to Resolution Protocol (ARP) but different in realization
. If a particular EUS wants to update its address translation table, this will
And sends a network message to a well-known address translation server. Then, corresponding
MAN address is returned. Provide a set of such address translation services
By offering N is a number of different, new and existing network software within the Sun environment
It can function as a sub-network for services.8.3.3.2.2 Device Driver   On the upper side, the driver has a higher protocol and protocol for transmission.
Multiplexes a plurality of different queues of LUWUs from the
Queue up the UWU to several different queues for higher tiers. hard
On the wear side, the driver can write to the user memory buffer or
Set up the upcoming DMA transmission. The driver saves the user buffer.
Can be accessed by the DMA controller through the in-system bus
Communication with the system is required to map into a possible memory.   In transmission, the driver uses a protocol layer that does not use the MAN address.
In other words, it is necessary to translate the output LUWU address for the ARPAnet protocol.
It is necessary. MAN destination address and MAN datagram control sent when destination group transmits LUWU
In the information. Other transmission control information includes LUWU length,
Columns indicating service type and higher level protocol, and DMA
The data location for this is included. The UIM uses this control information to send a packet
Form a heading and send the LUWU data from the EUS memory.   In reception, the driver enters the poll / response protocol into EUS and the data
Is realized by the UIM that notifies the user. This poll response contains the originating address and the LUWU
Indicates overall length, amount of data arriving to date, higher protocol layer
Includes a type field, and some agreed upon amount of data from the message. (
For small messages, this poll response must contain the entire user message.
You can also). How the driver itself interprets this message based on the type field
Flexibility to receive and determine to which higher level entity to pass this
Have Based on type field This could mean simply delivering the notification and passing the reception decision to a higher tier.
You. Whatever approach was used, the UIM then issued an EUS note.
Control request to deliver the data to the
A DMA operation by the IM is performed. EUS buffer for receiving data
Pre-assigned to protocol type, driver fixed reception
Can be handled in a pre-determined format, or simply as a notification.
It is possible for Eva to obtain buffer information from a higher layer and handle it.
This allows the driver to handle both existing and new applications in the Sun environment.
Is the type of flexibility required to8.3.3.3.3 Raw MAN interface software   Application for the direct use of MAN network functions in the future
  When the program was created, this address translation function was necessary. Disappears. MAN datagram control information is stored in dedicated MAN network layer software.
Can be specified directly by the software.9. MAN protocol 9.1 Overview   The MAN protocol is a user from the originating UIM to the destination UIM across the network.
Perform data delivery. This protocol is connectionless,
Asymmetric with respect to transmission and reception, detecting errors but not correcting
To achieve the purity of the layer to achieve.9.2 Message scenario   EUS sends datagram transactions, called LUWUs, into the network.
Data coming from the UES resides in the EUS memory. Control message from EUS
Is the data length for the UIM, the destination address for this LUWU, and the destination group.
Loops and information such as the user protocol and the network class of the required service.
Specify the type field that contains the report. Together, this data and control information Form an LUWU. Depending on the type of EUS interface, this data and
And control information is passed to the UIM in different ways, except that data is
There is a high possibility that it will be passed by sending.   The UIM sends this LUWU to the network. Large LUs to reduce potential delays
WUs are not sent to the network as one continuous stream. UIM has LUWU
Cut into pieces called packets that can have a certain maximum size. This
UWUs smaller than the large size are called SUWUs and are contained in a single packet
. Multiple EUSs are collected at NIM and these packets are sent from UIM to NIM
Link (EUSL). A packet from a certain UIM is
Request multiplexing (demand) with a packet from another EUS on the link (XL) to the INT
 multiplex). Delay, EUS all share link to MINT
Low because there is no need to wait for the end of a long LUWU transmission from another EUS.
Is reduced. The UIM uses the original LUWU trap for each packet. Generates informational headings from transactions, which allow individual packets
The packet is passed from the originating UIM to the destination UIM over the network, and
To the same LUWU that was passed to the network. This packet header is
Contains information on network layer protocols in the MAN network.   Before NIM sends the packet to MINT on its XL, this is NIM / MIN
Add a T-header to this packet message. This heading is specific EUS / U
Contains the outgoing port number identifying the physical port on the NIM to which the IM is connected. this
The heading is sent by MINT and the EUS is located at the port for which the user has permission.
Used to verify or place. One or more additional checks of this type
Secures the privacy of this temporary network depending on the data network that handles multiple temporary networks
Is especially important for you. MINT uses this packet header for the packet.
Used for determining routes, as well as for other possible services. MINT does not change the contents of the packet header. ILH in MINT
When passing through the network to send on XL to the destination NIM, this is NIM / M
Put a different port number in the INT heading. This port number is the destination EUS / UI
M is the physical port on the connected NIM. The destination NIM uses this port number.
This packet is sent to the corresponding EUSL on the fly.   Various sections in the packet are delimited according to the link format
Identified by This delimiter is NIM / MINT heading 600 and MAN heading
610 and between the MAN header and the rest of the packet. M
The delimiter at the boundary between the AN header and the rest of the packet is the header inspection sequence.
Required to insert or check heading checks in the sense circuit.
Required. NIM converts the received packet to all NIM / MINT header fields.
Broadcast to port.   When a packet arrives at the destination UIM, it The packet header is the outgoing U that is needed to reassemble the outgoing EUS transaction.
Contains the original information from the IM. In addition, pipeline connections, or other different
Includes various EUS transaction size, priority, and preemption schemes
Required to enable various EUS receiving interface approaches
Includes sufficient information.9.3 Description of MAN protocol 9.3.1 Link layer functions   The link function is described in section 5. Start of message
And end separators, data transparency, message checking on EUSL and XL links
The function of the lock sequence is discussed here.   Check sequence for entire packet message at link level
It is performed in. However, here, instead of performing the corrective action,
Are passed to the network layer to be processed here. Message
If there is an error in the check sequence, The error count is incremented for purposes and message transmission continues. Separate look
The outgoing check sequence is calculated in hardware in the UIM. MINT
If the control detects a heading check sequence error,
As a result, the message is discarded and the error count is incremented for administrative purposes.
This also occurs when there is an error in the header check sequence in the destination UIM.
Messages are discarded. The result of the data check sequence is LUWU
Delivered to the EUS as part of the delivery notification, and whether the UES receives this message
Can be determined. These violations of bed purity increase speed and overall net performance.
This is performed to reduce processing in the link layer.   Other "standard" link layer features such as error correction and flow control
It is not performed by conventional methods. Error correction at link level (retransmission request)
Or no notification message for flow control is returned. Florco
Account Rolls are signaled using dedicated bits in the framing pattern. X. 25
The complexity of protocols such as
Tolerant of high-speed links, improving the reliability of links with high error rates
. However, the low bit error rate of the fiber optic links in this network is unacceptable.
Is expected to achieve some level of error-free throughput (this
The bit error rate of the fiber optic link is less than 10 errors / trillion bits).
Also, non-communications in MINT and UIM required to process data from high speed links
Due to the always large amount of buffer memory, flow control messages
Deemed necessary or ineffective.9.3.2 Network layer 9.3.2.1 Function   The message unit that leaves the originating UIM and travels to the destination UIM is a packet
is there. This packet can be changed once it leaves the outgoing UIM There is no.   The information in the UIM-to-UIM message header allows the following functions to be performed:
I do.   Fragmentation of the LUWU at the originating UIM,   -Re-association of the LUWU at the destination UIM;   Routing to the correct NIM at MINT,   Routing to the correct UIM / EUS port at the destination NIM,   -MI of variable length messages (eg SUWU, packets, n packets)
NT transmission,   -Congestion control and arrival notification of the destination UIM,   Detection and handling of message header errors;   Addressing of network entities for intra-network messages;   UES authentication to deliver network services only to authorized users.9.3.2.2 Format   FIG. 20 shows a message format from UIM to MINT. MAN
Out 610 is the destination Address 612, source address 614, group (virtual network) identifier 616, group
Group name 618, service type 620, packet length (data included in header
Bytes) 622, service indicator type 623, end-user system
System used to identify the EUS to EUS heading 630 by the system.
Protocol identifier 624 and a header check sequence 626. This heading
It has a fixed length and is seven 32-bit words, or 224 bits long. M
EUS or EUS header 630 to fragment messages into AN headers
Followed by The heading includes an LUWU identifier 632, a LUWU length indicator 634,
EUS which is the header of packet sequence number 636 and user data 640
Protocol identifier 638 for identifying the contents of the internal protocol, and LUW
It contains the first byte number 369 of the data of this packet in all information of U. And most
Later, following destination port identification 642 and source port identification 644, the appropriate
To the protocol The corresponding user data 640 is sent. This column has 32 bits,
, For current network control processors, this is the most efficient length (integer)
That's why. Error checking is performed for this heading within the control software.
This is called the heading check sequence. At link level
Error checking is performed on the entire message,
Check sequence 634. For the purpose of completion, the drawings
MIN / MINT heading 600 is also shown.   The destination address, group identification, service type, and source address are MIN
Placed in the first five fields of the message for T processing efficiency. Destination and guru
Loop identification is used for routing and size is used for message management.
Type column for special handling and outgoing column for service awareness
Used for9.3.2.2.2.1 Destination address   The destination address 612 indicates to which EUS the packet is This is a MAN address that specifies whether a message is being sent. MAN address is 32
This is a flat address that specifies the EUS connected to the network. (
For intra-network messages, the high order bit in the MAN address is set.
If so, this address is used instead of EUS instead of an entity in the network, for example, MI.
Specify NT or NIM). MAN address is permanently assigned to a certain EUS
This EUS is identified even when it moves to a different physical location in the network.
When the EUS moves, it is signed by a well-known routing authentication server and its M
It is necessary to update the correspondence between the AN address and the physical port where it is located
is there. Of course, the port number is supplied by the NIM, so the EUS
You can't lie about it.   Within MINT, the destination address is used to route the message.
Used to determine destination NIM. In the destination NIM, this destination address
Less routes messages Used to determine the destination UIM.9.3.2.2.2.2 Packet length   The packet length field 622 is 16 bits long and contains a fixed header for this message fragment.
Indicates the length of bytes including data and data. This length is set by MINT
Used to transmit messages. This is also the data delivered to the EUS by the destination UIM.
Used to determine how much data there is.9.3.2.2.2.3 Type column   The service type field 623 is 16 bits long and is specified in the original EUS request.
Includes the type of service being serviced. MINT looks up the type of this service and
Changes the way messages are processed. The destination UIM is also this service Thailand
Check the destination and how to deliver this message to the destination EUS,
To determine whether or not the item should be delivered even if it exists. User protocol 624
The EUS driver in multiplexing various data streams from
.9.3.2.4 Packet sequence number   Described here is the packet sequence for this particular LUWU transmission
Number 636. This helps to re-join the incoming LUWU with the receiving UIM.
That is, the receiving UIM knows if the transmission fragment was lost due to an error.
it can. The sequence number is incremented for each piece of the LUWU. Last
The sequence number is negative, which indicates the last packet of the LUWU.
(One SUWU has -1 as a sequence number). LUW of infinite length
When U is transmitted, the packet sequence number is wrapped around.
It is. (For a description of LUWU of infinite length, see UWU Length, Section 9.3.
. See 2.2.7. )9.3.2.2.2.5 Outgoing address   The outgoing address 614 is 32 bits long, which indicates the E that sent the message.
MAN address specifying US. (For the description of the MAN address,
Destination address). This source address is stored in MINT for network accounting.
Hey Required. This along with the port number 600 from the NIM / MINT heading
It is used by MINT to certify outgoing EUS for network services
. The originating address is the destination EUS's network address of the EUS that sent the message.
Sent so that you can know the news.9.3.2.2.2.6 UWU ID   UWU ID 632 is used by the destination UIM to reassociate the UWU.
32 bit number. This reassembly is done by changing the order of the fragments in the network.
Note that it can be easier because it is not. The UEU ID is the source and destination address.
With the same LUWU, that is, the original datagram
Identify fragments of the action. This ID is transmitted while any fragment is in the network and
Must be unique for the destination pair.9.3.2.2.2.7 UWU length   The UWU length 634 is 32 bits long, and the entire length of UWU data is expressed in bytes.
Shown. LUWU In the first packet of this, the destination UIM performs congestion control
If the LUWU is pipelined to the EUS, this will
When the IM starts notification of the LUWU and before the entire LUWU reaches the UIM,
Can be delivered.   Negative lengths allow infinite LUWUs, such as an open channel between two EUSs
Indefinite LUWU close-down indicates sending negative packet sequence number
It is done by doing. When UIM controls DMA to EUS memory
In that case, only infinite length LUWU makes sense.9.3.2.2.2.8 Heading check sequence   There is a header check sequence 626, which is sent by the sending UIM
Information, so that the MINT and the destination UIM have correct header information.
It can always be determined whether or not it has been transmitted. The MINT or the destination UIM is
If there is an error in the sequence, do not deliver the packet. 9.3.2.2.9 User data   The user data 640 contains the user UWU data transmitted in this transmission fragment.
Data. Message calculated at link level with this data
The entire inspection sequence 646 follows.9.3.3 NIM / MINT layer 9.3.3.1 Function   This protocol layer consists of a heading containing the NIM port number 600. This port
The port number has a one-to-one correspondence to the EUS connection on the NIM,
In block 403 (FIG. 16) the user inserts fake data into this
Is generated so that it cannot be done. This heading is placed before the packet message,
Overroll packet not covered by message inspection sequence
No. This is because parity bits within the same word are
Checked by group. The incoming message to MINT is the outgoing NIM
Contains the network number for the network service required in the type field. Used for the authentication. Is outgoing message from MINT the outgoing port number 600?
Instead, the speed of demultiplexing / routing to destination EUS by NIM
To include the destination NIM port number. The packet is in one NIM
If you have multiple destination ports, the list of these ports is placed at the beginning of the packet.
The heading section 600 is several words long.10. Login procedure and virtual network 10.1 General   Systems such as MAN typically provide services to a large number of customers.
Most cost-effective. Many of these customers have external security
It is likely that multiple users will require protection. These users can access the virtual network.
It is convenient to make a loop. To provide greater flexibility and protection
To this end, individual users can be given access to multiple virtual networks. For example
Put all users of one company on one virtual network and payroll department of this company
A separate provisional It can be placed on the imaginary net. Payroll users may also be interested in
Belonging to both of these virtual networks requires access to general data
Required but outside payroll department users access payroll records
Must not belong to the virtual network members of the payroll virtual network.
Required.   The log-in procedure and routing for caller checking are performed by the MAN system using multiple temporary
Network to provide optimal levels of protection against unauthorized data access
It is a method conceived to make it possible to In addition, NIM will
This method of generating user port numbers for packets
Thus, additional protection is provided against access to the virtual network by unauthorized users.10.2 Authorization database construction   FIG. 15 shows the management control of the MAN network. Database is disk 3
Operating, management and maintenance (OA & M) system stored in 51 Access to authorize the user in response to a login request.
For large MAN networks, the OA & M system 350 handles a large number of login requests.
It is also conceivable that the present invention is a distributed multiprocessor device for processing. This database
Cannot access the restricted virtual network if the user is not a member
Designed to be. This database is under the control of three types of superusers.
I will The first super-user is an employee of a carrier that provides MAN services.
This super user, referred to here as the level 1 super user, is usually an individual user
Assign the block MAN name consisting of the block numbers for the group, and
Assign some specific of these names to Type 2 and Type 3 superusers.
Level 1 super users also assign virtual networks to specific MAN groups. Most
Later, level 1 super-users will be offered services provided by the MAN, such as electronic
Authorized to create or destroy “yellow pages” services. Type 2 or more
User assigned A valid MAN name from the allocated block to a specific user population and
If so, specify physical port access restrictions. In addition to this, Type 2 super users
, Restricting access to virtual networks with members of his set of customer groups
Have.   Type 3 super-users have roughly the same privileges as type 2 super-users,
Has the authority to permit access to the virtual network for the MAN name. Such an action
Seth sends a MAN name type 2 super user to this MAN name user in table 37.
If you allow the ability to join this group with an appropriate item in 0, type
Note that only 3 superusers are allowed.   The database includes a table 360 that includes individual user identifications 36.
2. Password 361, group 36 accessible using this password
3. List of ports from which the user will send and / or receive and special
Directory number 364, service type 365,
, Receive only, send A column indicating dedicated or receiving and transmitting is included.   This database also stores users (table 370) for individual users.
User to associate with a potentially allowed groove (table 375)
Capability tables 370, 375 are included. A user is a group by a superuser
If you want to be granted access to the
Is checked to see if it is in the list of table 370, and is not in the list
If this request to allow the user for that group is denied
It is. Superusers are against their groups, and those in tables 370, 375
Has the right to enter data for the group. Super users are also their users
Group from the user / group permission table 360
Group 363. Therefore, a certain you
In order for the user to access the outer group, both groups must
Who must allow this access. 10.3 Login procedure   At login, a user who has been given valid permission in the manner described above
The user sends an initial login request message to the MAN network. This message is
It is not directed to the user, but to the MAN network itself. In fact, this message
A message is a headline-only message that is separated by the MINT central control.
Is analyzed. Password, type of login service required, MAN glue
MAN name and MAN name of the login request are replaced with other fields.
Included in the voucher. This is only the heading of the MINT Central Control by XLH
Has been passed to be further processed by the OA & M central control
It is. MAN name, required MAN group name (virtual network name), and password
The login data including the password is compared with the login permission database 351 and the
A particular user has access to this virtual network from the physical port to which that user is connected.
It is checked whether access is allowed. (This thing The management port is pre-configured by NIM before receiving the login packet by MINT.
Will be undecided. ). If this user is in fact legitimately authorized, call
The tables in the checker 307 and the router 309 (FIG. 14) are updated.
Outgoing checker table for checkers that handle this logged-in user's port
Only the file is updated from the login for the terminal operation. Login request is received
The routing table of all MINTs responds to the request.
Any allowed connection of the same group connected to another MINT to answer
Must be updated to allow callers to receive data from competent users
. The outgoing checker table 308 stores data in the XLH for that outgoing checker.
Allowed names / groups for individual ports connected to the NIM sending the data stream
  Contains a list of pairs. The router table 310 is all receiving UWU
Contains items for all authorized users. Individual items are named / group pairs.
A and the corresponding N Includes IM and port number. Items in outgoing checker list are group identifiers
Grouped by number. Group identifier number 616 is the login user
Is part of the header of the subsequent packet from OA & M at login
Derived by system 350 and MAN switch 1 by OA & M system
0 is redirected to the logged-in user. OA & M System 350 is MIN
Log using access 19 to MINT memory 18 of T central control 20
Enter a login notification for the in-user. For subsequent packets,
When they are received in the MINT, the outgoing checker checks the port number, MAN name and
Check the MAN group name against the permission table in the outgoing checker,
As a result, it is determined whether the packet should be processed. Router is virtual next
Checking the network group name and destination name ensures that the destination is
To see if it is allowed. As a result, the user is logged in once
And the user You can access all destinations in the table. In other words, read-only mode
Previously logged in for access in read or read / write mode
Destination and destination with the same virtual network group name as required in the login
Can access all packets, while unauthorized users can block all packets.
Clicked.   In this embodiment, the check is performed on individual packets,
Do this for each work unit (LUWU or SUWU)
All subsequent packets of the LUWU whose original packet was denied are denied
It is also possible to record the indicator, or that the original packet
, All LUWUs that are lost at the location may be rejected.   For super-user logins related to login database changes, this is
Database stored on disk 351 in A & M system 350
Conventional login except that it is recognized as a login request with the authority to change
N Is checked as well.   Super User Types 2 and 3 provide MAN access to OA & M system 350
Obtained from a computer connected to the user port. OA & M System 350
Derive statistics on billing, usage, permissions and performance, but this
Thus they can be accessed from their computer.   MAN networks also provide special types of users, such as transmit-only users and receive-only users.
We can also provide services to users. Broadcast as an example of a send-only user
Broadcast stock quotation
system) or a video transmitter. The output of the sending-only user is
Only checked in car tables. Receiving-only unit, e.g.
The linter or monitor device is identified by an entry in the routing table.
Allowed.11. Application as MAN voice switch   Fig. 22 shows the audio sequence of the MAN architecture. Shows a configuration for use to switch data. This architecture
Existing switches to simplify the application of these services to
In this case, it is manufactured by AT & A Network Syeteme.
BuiltThe advantage of using a switch is that it requires a lot of development effort.
By eliminating the need to develop programs to control the Cal switch
is there. Use existing switches as interfaces between MAN and voice users
By doing so, this effort can be almost completely eliminated. Fig. 22 shows 5ESS
A conventional customer telephone connected to the exchange module 1207 of the switch 1200 is shown.
Is done. This customer phone is also an integrated service digital network.
 digital network, ISDN) and 5ESS switch
And data customer stations. Other customers
Station 1202 is a replacement module 1207 Is connected through a subscriber loop carrier system 1203 connected to.
Switching module 1207 is a time division multiplex switch that establishes a connection between switching modules.
H1209. These two switching modules use common channel signaling
7 (CCS7) signaling channel 1211, pulse code modulation (PCM) channel 1
213 and an interface 121 comprising a special signaling channel 1215
Connected to 0. These channels are for interface with MAN NIM2
Is connected to the packet assembler and disassembler 1217. PAD
The functions are PCM signals generated in the switch and packet signals exchanged in the MAN network.
Interface with the issue. Special Signaling Channel 12
Fifteen functions allow the PAD 1217 to provide individual PCM channels and associated callers and destinations.
To inform about CC7 channel sends packet to PAD 1217
However, the PAD 1217 converts this packet into the format required for exchange by the MAN network.
Processing for Do. To make the system resistant to equipment or transmission facility failures,
This switch is connected to two different NIM networks of the MAN network. Digital PBX1
219 also interfaces directly with the packet assembler disassembler 1217.
Face. If you want to upgrade the PAD later,
An integrated service digital that directly generates digital audio bit streams
Can interface directly with telephones such as network (ISDN) telephones
It is.   The NIM is connected directly to the MAN hub 1230. NIM is the MIN of this hub
Connected to T11. MINT 11 is interconnected by MAN switch 22
You.   For this type of configuration, to make the most efficient use of the MAN hub,
Significant amounts of data as well as voice are required to be switched. Voice packet
In particular, minimizes the total delay encountered when transmitting audio from a source to a destination.
And loss of some of the audio signal. Very short delays to ensure that there are no large inter-packet gaps.
Has an extension element.   Basic design parameters for the MAN are selected to optimize data exchange
And applied in the simplest way as shown in FIG. Large amount
If more than one voice packet exchange is required, one or more of the following additional steps:
Is taken.   1. Provides excellent performance in encoding forms, eg, 32 Kbit / s
Adaptive Differential PCM (ADPCM) is used instead of 64K bit PCM. Performance
Better coding schemes that require bit rates of 32 bits per second or less to improve
Chimes are provided.   2. Packets are required to be sent only when the customer is actually talking
. This reduces the number of packets to send to at least 2: 1.   3. The size of the buffer for buffering audio samples is 256 audio samples (
More than memory for 2 packet buffers) / channel You can also. However, longer voice packets introduce more delay, which is
This depends on the characteristics of the rest of the voice network.   4. Switch the voice traffic to switch setup operation for voice packets.
It is also possible to concentrate in a specialist MINT in order to reduce the number. However
Such a configuration increases the number of customers affected by NIM or MINT failures.
Or provide an alternative route to another NIM and / or MINT
It is also conceivable that a configuration for this is required.   5. Other hub configurations can be used.   The alternative hub configuration shown in FIG. 24 is an example of the solution of step 5. Voice packet
The basic problem in exchanging packets is to minimize delay in voice transmission.
Audio packets must be represented by short segments of audio
That is, according to some guesses, this length is as short as 20 milliseconds
. This translates into a large number of 50 packets per second for each direction of voice. Equivalent to. If a significant amount of input to MINT is such a voice packet
Circuit switch setup time is too large to handle this traffic
There is danger. High traffic situations where only voice traffic is exchanged
Requires a packet switch that does not require circuit setup operations
It is also possible.   One embodiment of such a packet switch 1300 is a space division switch.
Consists of a group of MINTs interconnected like a conventional array of
, The individual MINTs 1313 are connected to the other four, and all the output MINTs 1312 are connected.
Are added to carry a sufficient amount of voice traffic to reach. Because of the device
For additional protection against failures, the MINT 131 of the packet switch 1300
3 are interconnected through the MANS 10 and the MINT 1313 that has failed the traffic
And a spare MINT1313 can be used instead.
You.   The output bit stream of NIM2 is Connected to one of the inputs (XL). Packet data leaving input MINT 1311
Data traffic can subsequently be switched to the MANS10. This implementation
In the embodiment, the data packet output of the MANS 10 receives the output of the MANS 10.
Output together with the voice packet output of the data switch 1300 in the MINT 1312
Are combined. The output MINT 1312 is connected to the MANS 10 on the XL16 (input) side and the data.
The IL17 output is received by the PASC circuit 290.
13 is an input bit stream of NIM2 generated by (FIG. 13). Input MINT
Reference numeral 1311 denotes a PASC circuit 290 (No. 13) for generating an output bit stream to NIM2.
13). For output MINT1312, X from MANS10
The input to L can be many different through the circuit path where this input inserts different delays.
Phase matching circuit 292 (FIG. 13) as shown in FIG. 23 because it comes from the source
Passed.   This configuration also transmits high priority data packets. Pass the low-priority data to the circuit switch 10
  It can also be used to hold packets to exchange. In this configuration
The packet switch 1300 does not carry voice traffic.
No need to pass 312, in this case to MINT that does not carry voice traffic
Is required to be directed to the circuit switch MANSI0. 12. MINT access control to MAN exchange control   FIG. 21 controls the access of the MINT 11 to the MN exchange control 22.
The configuration for this is shown. Each MINT has one associated access controller 1
It has 120. Data rings 1102, 1104, and 1106 have individual access
Controller free logic and output link to count circuit 1100
Distribute the data shown. Each access controller 1120 has data
Holds an output link that wants to send
, Each link includes an associated priority indicator 1114. MINT uses this resource
Mark the output link of the strike as busy in the ring 1102;
Output link required from MAN exchange control 22 from ILH of this MINT
Set route to It can be captured by sending an order to be up. Its output
When all of the data blocks to be transmitted to the
Output links are identified as empty in the data transmitted by data ring 1102.
The output link can be accessed by other MINTs.
To do.   One problem with using only the availability data is that when traffic jams occur,
MINT takes longer to get access to one output link
It is too much. Access to output links to MINT can be averaged
, The following configuration is used. Ready bits (r) transmitted in link 1102
eady bit) associated with the individual link availability indicator
Window bits exist. This ready bit is output
Controlled by any MINT that captures or releases the link. This win
The dow bit is a single MINT, the control MINT for the purposes of this description. Controlled by MINT-only access controller 1120
. In this particular embodiment, the control MINT for any output link is paired.
The corresponding output link is the MINT directed to it.   Open window (window bit = 1) is the output link on the ring
Wants to be captured, and the ready bit passes
The first access controller, which recognizes that it is free, seizes this link.
Forgive and prevent any controller from attempting to capture the link in use.
Allow setting priority indicator 1114 for busy links. Black
Window (window bit = 0) for the corresponding free link
This free link only to controllers with a priority indicator set to
Allow to capture. This window is used to access the MINT
1120 causes the logic and count circuit 1100 of the controller to output.
While the power link is in use (Ready bit = 0), the controller is closed.
When it detects that the power link is free (ready bit = 1),
Is performed.   The link capture operation of the access controller is as follows. Used by link
Medium (ready bit = 0) and the window bit is 1,
Access controller sets a priority indicator 1114 for that output link.
To If the link is in use and the window bit is zero,
Do nothing. When the link is idle and the window bit is 1
The controller will seize the link and no other controller will seize the same link.
Mark the ready bit as zero. Link is free, window
If the bit is zero, the priority indicator 1114 is set for that link.
Only the installed controller can capture the link and the ready
Mark the bit as zero to capture it. Window to bit The operation of the access controller of the control MINT is simple. In other words, this
Some controllers simply copy the value of the ready bit into the window bit
You.   In addition to the ready and window bits, the frame bits are
6 defines the start of a frame of resource usage data, and therefore individual release and
To define a count to identify the link associated with the window bit
Will be patrolled. The data on the three rings 1102, 1104 and 1106 is serial
Cycles through the logic and count circuit 1100 of each MINT synchronously and synchronously
I do.   An attempt to capture one output link as a result of this type of operation, and
And the unit and window that first successfully captured this output link
Access control located between the access controller that controls the bits
Is given priority and then makes a request to seize this particular output link.
Other controllers Is processed before As a result, all MINTs are
A fair distribution of access is achieved.   This is used to control MINT11 access control to MANSC22.
Is controlled from the preferred MINT if the alternative approach is used. Individual M
INT maintains priority and normal queues for queuing requests, and
A request for a service is first made from the MINT priority queue. 13. Conclusion   The above description is merely related to one preferred embodiment of the present invention.
Obviously, many other configurations can be designed without departing from the spirit and scope of the
The present invention is limited only by the claims.               List of code names 1SC first stage controller 2SC 2nd stage controller ACK notification response ARP address resolution protocol ARQ automatic repeat request BNAK busy negative notification CC central control Negative response of CNAK control CNet control network CRC cyclic redundancy check or code DNet data network DRAM Dynamic Langum Access Memory DVMA direct virtual memory access EUS end-user system EUSL end user link (connect NIM and UIM) FEP preprocessor FIFO FIFO FIFO FNAK Tissue Blocking Negative Notification IL internal link (connects MINT and MANS) ILH internal link handler IP internal protocol LAN local area network LUWU Long User Work Unit Metropolitan area network as an example of MAN MANS MAN switch MANSC MAN / Switch Controller MINT memory and interface module MMU memory management unit NAK negative notification NIM network interface module OA & M operation, management and maintenance PASC phase matching and scrambling circuit SCC switch control complex SUWU short user work unit TPC transmission control protocol TSA time slot allocator UDP User Datagram Protocol UIM user interface module UWM User Work Unit VLSI large-scale integrated circuit VME bus One IEEE standard bus WAN wide area network XL external link (connect NIM to MINT) XLH external link handler XPC crosspoint controller

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 shows graphically the characteristics of communication traffic of the type found in a metropolitan area network; FIG. 2 shows an example including a typical input user station communicating over this network FIG. 3 shows a high-level block diagram of a metropolitan area network (referred to herein as a MAN) as an example; FIG. 3 is a more detailed block diagram of a MAN hub and units communicating with the hub; FIGS. FIG. 6 is a block diagram of a MAN showing how data moves from an input user system to a MAN hub and then to an output user system; FIG. 6 shows an example of a type that can be used as a circuit switch in the MAN hub. FIG. 7 shows a block diagram of an exemplary embodiment of a MAN circuit switch and its associated control network. 8 and 9 are flowcharts showing the flow of requests from the data distribution stage of the hub to the controller of the circuit switch of the hub; FIG. 10 is a block diagram of one data distribution switch of the hub. 11 to 14 show a block diagram and a data layout of a portion of the data distribution switch of the hub; FIG. 15 shows the operation, management and maintenance (OA) for controlling the data distribution stage of the hub.
& M) a block diagram of the system; FIG. 16 is a block diagram of an interface module for the interface between the end user system and the hub; FIG. 17 is a block diagram of the interface between the end user system and the network interface; FIG. 18 is a block diagram of a typical end user system; FIG. 19 is a block diagram of a control device for the interface between the end user system and the hub of the MAN. FIG. 20 is a layout of a data packet designed for transmission over the MAN to illustrate the MAN protocol; FIG. 21 is another layout for controlling access from the data distribution switch to the circuit switch control; FIG. 22 shows a MAN with audio and data. FIG. 23 is a block diagram illustrating an arrangement for use to exchange data; FIG. 23 illustrates an apparatus for synchronizing data received from a circuit switch by one of the data distribution switches; FIG. 25 shows another configuration for a hub for exchanging structured voice and data; and FIG. 25 is a block diagram of a MAN circuit switch controller. [Description of Signs of Main Parts] 2 Network interface module 10 MAN switch 11 Interface module 12 Internal link 13 User interface module

Claims (1)

[Claims] 1. A method for ensuring security in transmitting a packet from an input port to an output port, the method comprising: identifying the input port transmitting the individual data packet within the individual data packet; and identifying the user of the input port. And checking for a pre-authorized pair of the user identification and the port identification for the individual data packet. 2. 2. The method according to claim 1, wherein the identification of the port is provided by the data network and is not free to the user at the port. 3. In a data network for transmitting data packets, the network comprises: means for inserting in a packet the identification of the port transmitting the packet, the network being located in the network and depending on the user at the port; Uncontrollable means; and means for authenticating from the port identification and addressing data in the packet whether the port is authorized to transmit the packet to the network. network. 4. 4. The data network of claim 3, wherein the means for authentication further authenticates whether the port is authorized to transmit the packet to a destination user identified in the addressing data. A data network comprising means for: 5. A method used in a data network to secure a secure transmission from a source user to a destination user, the method comprising: providing the destination user with a destination user password, destination user identification, destination group identification and Logging in to the system with a login data packet comprising destination port identification; the data network receiving a packet for the destination group and destination user with the destination user password, destination user identification, destination user group number and destination user port number. Authenticating the user as authorized to do so; the source user logging in with a login packet comprising the source user identification, source user password, source group identification, and source port identification provided by the network to the system. Step to do Authenticating the source user password, source user identification, source user group identification, and source user port identification; recording authorization for the identification of the source user, source group, and source port in a source table; routing table Recording the authorization for the identification of the destination user and the destination group and the destination port in the source user identification and source group provided in the source table by the network for each transmitted packet. Checking the identification and source port identification and finding the destination port using the destination user identification and destination group identification in the routing table; and the source and destination port finding steps by the source check and destination port discovery steps. When the the destination is recorded in the source table and the destination table is confirmed, a method which comprises the step of transmitting the packet to a destination port identified in said finding step. 6. 6. The method according to claim 5, wherein the source and destination groups for one transmitted packet are the same, so that only users with a common group identification are allowed to communicate. . 7. 7. The method of claim 6, further comprising: discarding the packet if the result of the source check and destination discovery step indicates that the source and the destination are not recorded in the source table and the destination table. A method characterized by being performed. 8. The method of claim 7, further comprising recording source and destination data for the packet to be discarded. 9. 7. The method according to claim 6, wherein the step of discovering destinations further comprises the step of inserting an identification of the destination port found in the step of discovering destinations into each transmitted packet. 10 ． In a device in a packetized data network, the device identifies the source network port provided by the network and the source and destination names provided by a source user system. Apparatus comprising means for inserting into a packet header, and means for checking whether the source name is authorized for transmission from the port to the destination name. 11 ． An apparatus in a packetized data network, comprising: identifying a source port; identifying a source user; identifying a destination user system; identifying a user group; identifying a type of service to be provided; Means for inserting a header check for detecting an error in the network header into a data packet header of the network; identifying the source port; and identifying the source port as identification of the source port to the network header. Means for inserting; means for checking whether the combination of the source user, the user group and the source port are authorized to transmit data packets on the data network; Means for generating a destination port identification from the combination of groups and transmitting the data packet to the destination port An apparatus characterized in that: 12 ． The apparatus of claim 11, further comprising: an indication of a length of the user work unit, an indication of a number of packet sequences in the user network, an indication of a protocol to be used by the destination user, and the user work unit. Means for inserting the number of the first byte of the packet into the header and user header data identifying the identity of the user work unit; and referring to the other packets of the user work unit, Means for identifying the packet using the packet sequence number; recognizing the transcendence of the sequence packet using the packet sequence number; and storing the user work unit using the number of the first byte. Means for storing data of the packet associated with the user, and a user using the length indicator Apparatus characterized by comprising a recognizing means the completion of reception of Kuyunitto. 13 ． A method for preventing one group of users from obtaining an unauthorized address to another group of users in a network for serving users of a plurality of user groups, the method comprising: Generating a user authorization database for allowing access to one member of a plurality of groups by a plurality of users; including identifying a login destination user group from the first user and identifying the first user. Processing a login packet to determine whether the first user is allowed to access the destination user group in the authorization database; When the first user indicates to be authorized to access the destination group, Recording data indicating that the packet is authorized to be transmitted; thereafter, each and every individual within the recorded data including the identification of the first user and the identification of a destination user group for the data. Confirming to the data entity whether the first user is authorized to transmit data to the destination user group; and if the confirming step indicates that the first user is authorized. Transmitting the data entity to users of the destination user group. 14 ． 14. The method according to claim 13, wherein the data entity is a packet. 15 ． 14. The method of claim 13, wherein the login packet and the data entity each include an identification of a user port; the first user if the recording step is a transmission from a user port identified in the login packet. Storing data indicating that access to the destination user group is permitted; wherein the confirming step identifies the first user within the data entity within the second recorded data. Verifying whether the data network from the user port is authorized; and providing a source user port identification for each of the data entities from the internal and external controls by the user. A method characterized by being performed. 16 ． 14. The method of claim 13, wherein if the authentication step indicates that the first user is authorized to receive data from the destination user group, the first user may receive a packet from the destination user group. Recording data indicating that the user is authorized to receive the user user port identification for each of the entities from inside and outside controls of the network by the user. Processing the login includes accessing a database containing a table of passwords and corresponding user and group identification; creating the user authentication database includes entering data to build an authentication table. The checking step for the individual data entity is Ascertaining whether the first user is authorized to transmit data to the destination user group for all data. 17 ． In a method for transmitting a data entity from a source user to a user that is a member of a group in a data network, the method comprises: for each data entity including an identification of the source user and the group; Verifying that the source user is authorized to transmit a data entity to one user of the group; and if the verifying step indicates that the source is authorized, the data entity belongs to the group. Transmitting to one user. 18 ． A method for preventing one group of users from gaining unauthorized access to another group of users, used in a network to serve users of a plurality of user groups, the method comprising: Generating a user authorization database for authorizing a first user to access one member of the plurality of groups; the first user transmitting the data to a data packet transmitted by the first user; Providing a source user port identification from controls inside and outside the network; identifying a destination user group from the first user; identifying the first user;
And processing a login packet including a user port identification for the first user, wherein the first user transfers a data packet from the source user port for the first user in the authentication database to the destination user group. Determining whether transmission to one member is permitted; the authentication step allowing the first user to transmit a packet to one member of the destination user group; If so, recording data indicating that the first user is authorized to transmit packets from the source user port to the destination user group; The first user indicates that he is authorized to receive data from a member of the user group. Is authorized to receive packets from the members of the destination user group at the source user port; then the identification of the first user, the user for the first user port,
And for an individual data packet including an identification of a destination user group, whether the first user is allowed to transmit data from the port identification of the data packet to one member of the destination user group Confirming; if the confirming indicates that the first user is authorized, transmitting the data packet to a user of the destination group; and identifying a user and a user port transmitting the unauthorized packet. Recording the. 19 . In a data network, the network comprises: a source authentication database containing data indicating authentication to the source user and a group of source users for each active source user; A destination authentication database containing data indicating authentication for the destination user group; and the source user and group are authorized to transmit to the destination user and group in response to data in data packets received by the network. A data network comprising means for checking whether or not the data network is valid.