JP3805884B2 - ATM relay device - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to an ATM relay device.
[0002]
[Prior art]
With the rapid spread of the Internet and intranets, the increase in communication volume and the rapid increase in demand for real-time information support such as voice and video means a demand for building a communication infrastructure that can process large volumes of information in real time at high speed. Yes. The core is the router. Up to now, router technology has realized packet transfer processing using software technology, and the number of packets that can be processed per unit time is limited and the time required for processing varies. Therefore, it cannot be said that it will always be able to cope with the large capacity and real-time communication required in the future. Since the performance of the conventional router device is determined by software processing, that is, the performance of the CPU, it is difficult to greatly improve the performance by introducing a new high-speed technology centered on such technology. A review from the hardware architecture of IP packet relay is forced. As a means for solving such a bottleneck with respect to the performance of the existing router, there is a technique that integrates ATM exchange technology and IP processing. Although the function is a router, the packet transfer speed is increased by the power of the ATM switch, that is, the software processing is not performed for each IP packet, but a direct ATM connection is used. A router using such an ATM technology is known as a CSR (Cell Switch Router).
[0003]
FIGS. 8A and 8B are diagrams showing a first example of a CSR configuration, in which an ATM I / F is used for CPU communication between the ATM switch unit 2 and the IP processing unit 1. The software of the IP processing unit 1 includes a CSR protocol 1a, a UDP (User Datagram Protocol) / IP (Internet Protocol) 1b, a BSD kernel 1c, an ATM driver 1d, and a serial driver 1e. FIGS. 8C and 8D show examples in which an asynchronous port such as Ethernet is used for CPU communication between the ATM switch unit 2 and the IP processing unit 1.
[0004]
The ATM driver 1 d is used for connection with the ATM switch unit 2. The CPU of the IP processing unit 1 gives instructions to set and release the VPI (Virtual Path Identifier) / VCI (Virtual Channel Identifier) of the ATM switch unit 2 for cut-through by communicating with the CPU of the ATM switch unit 2. . The CSR simultaneously provides a conventional packet transfer mode by software processing, that is, a hop-by-hop transfer mode, and a high-speed packet transfer mode by hardware using an ATM switch. As will be described later with reference to FIG. 9, when the CSR of this configuration is used, the operation is disconnected from the IP processing unit 1 in the middle of communication, and the ATM switch unit 2 performs cut-through to increase the transfer rate. Can do. The cut-through means that a shortcut is taken, that is, the input packet is not always transferred to the router unit, that is, the IP processing unit 1, but is relayed to the output port only by the ATM switch unit 2. The cut-through process is not visible to end terminals and relay ATM switches. The end terminal may transmit an IP packet to an adjacent CSR (IP router) as in the case of using an existing IP router network.
[0005]
FIG. 9 shows the basic operation principle of CSR. As shown in FIG. 9A, the first packet is transferred in the hop-by-hop transfer mode, the packet is reconstructed in the CSR, and the packet header is analyzed. The CSR uses the packet header information to transfer the packet to the next hop node (router or destination host) or network (router). For packet flows that are expected to have a long application duration, such as ftp, telnet, WWW, etc., as shown in FIG. 9B, the packet flow and cut-through transfer are performed according to the protocol (CSR protocol) between adjacent nodes. The cut-through transfer path is established by notifying the correspondence relationship of the path (ATM-VC). The determination of the cut-through transfer path establishment is made (triggered) using the TCP / UDP port number information or the packet length.
[0006]
After the cut-through transfer path is established, as shown in FIG. 9C, the packet is switched by using the ATM header information VPI / VCI without being reconstructed in the CSR. When a decrease in the packet flow amount of the application is detected, the CSR releases the cut-through transfer path using the CSR protocol as shown in FIG. After the cut-through transfer path is released, the mode transits to the hop-by-hop transfer mode (FIG. 9A). The CSR protocol is effective only between adjacent nodes, and the information / state exchanged by the CSR protocol is effective only between adjacent nodes. Therefore, it is sufficient that the information / state regarding cut-through packet transfer is synchronized only between adjacent nodes, and global synchronization between all nodes is not required. Periodically refresh information / status. Further, the cut-through transfer path establishment policy is a local judgment of each CSR.
[0007]
FIG. 10 is a diagram showing the operation principle of the CSR triggered by RSVP (Resource Reservation Protocol). The basic part is the same as in FIG. First, as shown in FIG. 10A, the CSR operates as a normal IP router that performs processing by IP processing. At this time, the default virtual channel VC is used. Here, as shown in FIG. 10 (b), when an RSVP request message is received from the data receiving side (CSR-3), CSR-2 and CSR-3 and new channel VC are triggered by this message. Set (VPI / VCI). Similarly, a new channel flows between CSR-1 and CSR-2. After exchanging VPI / VCI with each other in the inter-CSR protocol, data flows using a new channel. High-speed packet transfer, that is, cut-through transfer is performed by directly connecting the channel between CSR-1 and CSR-2 and the channel between CSR-2 and CSR-3 by ATM (FIG. 10C). When the decrease in the packet flow amount of the application is detected, the CSR releases the cut-through transfer path using the CSR protocol as shown in FIG.
[0008]
As described above, attempts to transfer audio and video over the Internet have now been attempted, and there is a rapidly increasing demand for communication quality control and provision of multiple QOS classes for the future. In the flow-driven CSR system that provides ATM VCC for each packet flow, a method of requesting and providing a QOS (Quality of Service) class by sending a control message to a network from a node like RSVP (control-driven ), A method (data driven) for determining the QOS class using the information of the application type is easily possible.
[0009]
In consideration of flexible correspondence, the flow-driven cut-through path is generated without using control packet information and a method for providing a path generated by transferring a control packet such as RSVP to the network side, that is, a control-driven path. The path, that is, the data driven path should be operated in dual mode.
[0010]
However, when a cut-through path is generated every time an application flow arrives, that is, when only a flow-driven type operation is performed, the number of VCs required for forming a cut-through path becomes very large when applied to a large-scale network. There is concern.
[0011]
However, this problem can be solved by using a method (topology driven) in which a network administrator prepares a cut-through path having different QOS classes in advance. That is, in topology driven, packets transferred to a subnet having the same address can be shared by the same cut-through path, that is, merging can be performed. That is, since the cut-through path can be generated based on the network topology information, the number of VCs required for packet transfer by cut-through can be greatly improved by this topology driven.
[0012]
However, in topology-only operation, when performing quality assurance, for example, bandwidth assurance, the bandwidth prepared between nodes needs to be a bandwidth required when the amount of traffic (data flow) is relatively large. In other words, it is necessary for the network administrator to prepare a cut-through path having different QOS classes in advance, resulting in wasted bandwidth.
[0013]
Therefore, the dual mode operation that supports both of the flow driven and topology driven features is optimal.
[0014]
FIG. 11A shows the flow-driven operation described above, FIG. 11B shows the topology-driven operation described above, and FIG. 11C shows a dual mode in which both modes are combined.
[0015]
In other words, if flow-driven is applied simultaneously to topology-driven, a bandwidth can be secured on demand for a packet flow that requires quality assurance. This makes it possible to reduce the bandwidth used in topology-driven as compared to the case where only topology-driven is applied.
[0016]
By the way, a packet flow that shares a communication resource, that is, a VC among a plurality of applications, that is, a topology-driven path, and a packet flow that assigns a communication resource, that is, a VC for each application, or a packet flow that requires different communication quality classes exist simultaneously. Therefore, it is necessary to provide different communication quality classes in topology-driven in order to satisfy such requirements in topology-only operation. It is necessary to be able to operate a dual mode that simplifies hardware functions by using different cut-through paths for each required communication quality class.
[0017]
A topology-driven CSR network can be realized based on either a VP base or a VC base. In the case of realization based on VP, the maximum number of paths that can be set from one node is 256 (2 to the 8th power), resulting in a network with poor expandability. On the other hand, when implemented on a VC basis, it is necessary to support a VC merging function that guarantees cell ordering at the AAL (ATM Adaptation Layer) 5 level at a node where a plurality of flows join the topology-driven network.
[0018]
[Problems to be solved by the invention]
However, there are the following problems in performing the VC merging described above. That is, in the VC merging function as shown in FIG. 12, when cells (A1 to A4) and (B1 to B4) along two flows are simply merged, the cells are shown in the figure after switching by the ATM switching function 3. As shown in the figure, the flow is mixed, and when viewed at the AAL5 level, the cell BOC (Begin of Cell) that should arrive at the beginning of the packet arrives twice, and as a result, the PDU (Protocol Data Unit) cannot be assembled successfully. was there.
[0019]
The ATM relay apparatus of the present invention has been made paying attention to such a problem, and its purpose is to cut each communication quality class by merging cells in units of PDUs according to the communication quality. It is an object of the present invention to provide an ATM relay apparatus having a dual mode that can selectively use through paths.
[0020]
[Means for Solving the Problems]
In order to achieve the above object, the ATM relay apparatus according to the first invention has a high-speed packet transfer mode by hardware using an ATM switch, and packets having the same transfer destination address have the same path. An ATM relay apparatus for sharing and transferring, storing means for storing packets having the same transfer destination address in units of PDUs in different queues according to the communication quality of each packet, and units of PDUs stored in different queues Transfer means for arranging the packets in the order of the communication quality level and transferring them using a single path.
[0021]
The ATM relay apparatus according to the second invention has a high-speed packet transfer mode by hardware using an ATM switch, and transfers packets sharing the same path for packets having the same transfer destination address. An apparatus comprising: a disassembling unit that disassembles each packet having the same transfer destination address for each cell; an assembling unit that assembles the disassembled cells into PDU units according to communication quality; A transfer unit configured to arrange packets in units of PDUs in order of communication quality levels and transfer the packets using a single path.
[0022]
The ATM relay device according to the third invention further comprises a hop-by-hop transfer mode in which the received packet is processed by software in the ATM relay device according to the first or second invention, and packet transfer is performed in this hop-by-hop transfer mode. First processing means for performing processing, and second processing means for managing the ATM switch provided so as to be communicable with the first processing means are further provided.
[0023]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
[0024]
In order to solve the above-described conventional problems, in the present embodiment, as shown in FIG. 1, when merging a plurality of flows into one VC (Virtual Connection), a function 4 in units of VC (PDU) is used for each PDU. To align and merge cells. Here, such cell alignment is realized by using two embodiments described below.
[0025]
FIG. 2 is a diagram showing the configuration of the first embodiment of the present invention.
[0026]
In FIG. 2, the CSR includes an ATM switch unit 10 and an IP processing unit 11 connected to the ATM switch unit 10 via an ATM-IF 12. The IP processing unit 11 includes a CPU 11a and an AAL5 unit 11b. The IP processing unit 11 is a part that processes a CSR protocol between adjacent nodes or an IP packet that is not cut-through, that is, a hop-by-hop process, by software (CPU). A plurality (n) of IP processing units 11 may be connected to improve the capability of hop-by-hop processing. n = 1, 2,. . . However, it is determined according to the number of I / F ports of the ATM switch unit 10. That is, at least one ATM I / F 12 is required at least between the IP processing unit 11 and the ATM switch unit 10. An AAL5 unit 13 is provided to perform CPU communication between the IP processing unit 11 and the ATM switch unit 10, CSR protocol processing, and IP forwarding processing. The ATM switch unit 10 has a large number of ATM I / Fs 12 for connecting a large number of devices (routers, terminals, ATM exchanges, etc.) that can be connected via the ATM I / F 12.
[0027]
The ATM switch unit 10 of this embodiment includes an ATM switch unit core 100 that queues VCs for each PDU by an ATM switch, that is, performs cell alignment by a function based on perVC queuing.
[0028]
The ATMI / F input side will be described below. The ATM switch unit core 100 includes an input processing unit 101 for processing input cells, an address conversion table (which is not essential) 102 for performing address conversion for processing inside the ATM switch, and for each perVC, that is, for each perPDU. In addition, in consideration of QOS support, the per-VC (PDU) scheduler unit 104 that performs scheduling to each output port in consideration of the priority of each VC or PDU provides an ATM switching function The ATM switching function 103, the cell buffer unit 105 for storing the received cell, and the control unit for storing or reading the cell, that is, the address, attribute (QOS, priority, etc.) and pointer control of the cell buffer Buffer address pointer table 106 and VC Alternatively, an address conversion table 102 'for exchanging VPI / VCI addresses for outputting cells output after being arranged in units of PDUs to the ATMI / F, and a table that can refer to information on whether or not to perform VC merge VPI / VCI-> Merged VC unit 107. The selected VPI / VCI-> Merged VC unit 107 may also have the function of the address conversion table 102 ′. The cells output processed by the output processing unit 108 are sent to the ATM I / F 14.
[0029]
Furthermore, although the AAL5 unit 13 is connected to the ATM switch unit core 100, basically, when the CPU 15 of the ATM switch unit 10 communicates with the CPU 11a of the IP processing unit 11, the ATM switch unit core 100 and the ATM I This is for connecting both CPUs through / F14. The CPU 15 of the ATM switch unit 10 is a device management CPU, and controls the ATM switch unit 10 in response to an instruction from the IP processing unit 11 in accordance with the CSR protocol. That is, the ATM switch unit core 100 is controlled such as VPI / VCI setting and release, and various tables described above.
[0030]
FIG. 3 is a diagram for explaining the operation of the first embodiment of the present invention. In this embodiment, in the perVC queuing scheduler 104 shown in FIG. 3C, VCs are queued for each PDU for each output port of the ATM switch, that is, cells are aligned by a shaping function based on perVC queuing.
[0031]
That is, when input to the AYM switching function 103, as shown in FIG. 3A, cells with different QOS (priority 1 and priority 2) are mixed, and a PDU cannot be assembled correctly. Therefore, different queues #A, #B, and #C are provided for each priority, and cells having the same priority are stored in each queue in order (FIG. 3D). Next, scheduling according to the priority for each QOS is performed, that is, the cell data of each queue is merged and aligned so that the queue with the highest priority is the first in PDU units (FIG. 3B). . Then, it is output from the perVC queuing scheduler 104 in the PDU order as shown in FIG.
[0032]
FIG. 4 shows a modification of the first embodiment, in which perVC queuing, that is, perPDU queuing, is performed intensively at one location of the ATM switching unit.
[0033]
The second embodiment of the present invention will be described below with reference to FIG.
[0034]
First, the function of the ATM switch core 200 will be described. The explanation will start from the ATM I / F input side. The function may be simpler than that of the ATM switch core 100 described with reference to FIG. The ATM switch unit core 200 includes an input processing unit 201 that processes input cells, an address conversion table (this is not essential) 202 that performs address conversion for processing inside the ATM switch, ATM switching, ATM switching + scheduling unit 204 for providing a scheduling function, cell buffer unit 205 for storing received cells, and a control unit for storing or reading the cells, that is, cell buffer addresses and attributes (QOS, priority, etc.) And a buffer address pointer table 206 for performing pointer control and the like, and an address conversion table 202 ′ for performing VPI / VCI address conversion for output to the ATM I / F 300. The cell data is output by the output processing unit 208 and then input to the ATM I / F 300.
[0035]
The output port of the ATM switch unit core 200 is connected to the ATMI / F 300. On the port that supports topology driven by the ATMI / F 300, an AAL5 unit 303 that provides a SAR (Segmentation and Reassembly) function, and a buffer unit 301, a buffer address pointer table 302 that performs address management of the buffer unit 301, an address conversion table 209 that performs address conversion, and a table SelectedVPI / VCI-> Merged VC unit 207 that can refer to information about whether or not to perform VC merge Consists of The selected VPI / VCI-> Merged VC unit 207 may share the function in the address conversion table 209.
[0036]
Note that a plurality (m) (m = 1, 2,...) Of the AAL5 unit 303 of this embodiment is connected to the output port of the ATM switch unit 20. The AAL5 unit 303 performs the following cell disassembly and assembly processing.
[0037]
Further, although the AAL5 unit 23 is connected to the ATM switch unit core 200, basically, when the CPU 25 of the ATM switch unit 20 communicates with the CPU 21a of the IP processing unit 21, the ATM switch unit core 200 and the ATM I This is for connecting both CPUs through / F24. The CPU 25 of the ATM switch unit 20 is a device management CPU, and controls the ATM switch unit 20 in response to an instruction from the IP processing unit 21 in accordance with the CSR protocol. That is, the ATM switch core 200 is controlled such as VPI / VCI setting and release, and the various tables described above.
[0038]
FIG. 6 is a diagram for explaining the operation of the second embodiment of the present invention. In this embodiment, by connecting AAL5 to the output port of the ATM switch, the CS-PDU is taken out and then disassembled and reassembled. This is equivalent to scheduling in units of PDUs, and cells can be aligned by such a method.
[0039]
That is, in FIG. 6, when input to the AYM switching + scheduling unit 204, cells having different QOS (priority 1, priority 2) are mixed as shown in FIG. Can't assemble. Therefore, the data is input to the AAL5 unit 303 and decomposed for each cell in the AAL5 decomposition unit 303a under the control of the CPU 304. Next, the cells disassembled in the AAL5 assembling unit 303b are assembled into packets in units of PDUs according to QOS (priority 1, priority 2). As a result, it becomes as shown in FIG. These are sorted and output in PDU units in descending order of priority.
[0040]
FIG. 7 is a diagram showing a modification of the second embodiment described above, in which an AAL5 unit 303 is provided at a specific port of an ATM switch.
[0041]
As described above, this embodiment supports topology driven as a CSR function. Then, when merging packets transferred to a subnet having the same address into the same cut-through path, VCs are queued for each output port of the ATM switch or concentrated for each PDU by the above-described method, and the shaping function is used. The cells are aligned. As another method, scheduling is performed in units of PDUs by connecting AAL5 to an output port or a specific port of an ATM switch, taking out a CS-PDU and then reassembling and reassembling. Therefore, it is possible to provide an ATM relay apparatus having a dual mode in which cut-through paths can be selectively used for each communication quality class.
[0042]
【The invention's effect】
According to the present invention, cells are arranged in units of PDUs according to communication quality, merged and transferred, so an ATM relay device having a dual mode that can use different cut-through paths for each communication quality class Can be provided.
[Brief description of the drawings]
FIG. 1 is a diagram for explaining a basic concept of the present invention.
FIG. 2 is a diagram showing a configuration of the first exemplary embodiment of the present invention.
FIG. 3 is a diagram for explaining the operation of the first embodiment.
FIG. 4 is a diagram for explaining the operation of a modification of the first embodiment.
FIG. 5 is a diagram showing a configuration of a second exemplary embodiment of the present invention.
FIG. 6 is a diagram for explaining the operation of the second embodiment.
FIG. 7 is a diagram for explaining the operation of a modification of the second embodiment.
FIG. 8 is a diagram illustrating a configuration example of a CSR.
FIG. 9 is a diagram for explaining the basic operation principle of CSR.
FIG. 10 is a diagram for explaining a CSR operation using RSVP as a trigger;
FIG. 11 is an explanatory diagram of CSR topology-driven, flow-driven, and dual mode supporting both.
FIG. 12 is a diagram for explaining a problem when simple VC merging is performed in an ATM switching unit.
[Explanation of symbols]
10 ... ATM switch part,
11 ... IP processing part,
12 ... ATM / IF,
13 ... 5 copies of AAL
14 ... ATM / IF,
15 ... CPU,
100 ... ATM switch core,
101 ... Input processing unit,
102, 102 '... address conversion table,
103 ... ATM switching function,
104 ... perVC (PDU) scheduler,
105: Cell buffer,
106: Buffer address pointer table,
107 ... Selected VPI / VCI-> Merged VC section,
108: Output processing unit.

Claims (3)

It is an ATM relay device that has a high-speed packet transfer mode by hardware using an ATM switch and transfers packets having the same transfer destination address while sharing the same path,
Storage means for storing packets having the same forwarding address in units of PDUs in different queues according to the communication quality of each packet;
Transfer means for sorting packets in units of PDUs stored in different queues in order of communication quality level and transferring them using a single path;
An ATM relay device comprising: It is an ATM relay device that has a high-speed packet transfer mode by hardware using an ATM switch and transfers packets having the same transfer destination address while sharing the same path,
Disassembling means for disassembling each packet having the same forwarding address for each cell;
Assembly means for assembling the disassembled cells in units of PDUs according to communication quality into packets;
Transfer means for sorting assembled PDU-unit packets in order of communication quality level and transferring them using a single path;
An ATM relay device comprising: The ATM further includes a hop-by-hop transfer mode for performing software processing on the received packet, the first processing means for performing packet transfer processing in the hop-by-hop transfer mode, and the ATM provided to be able to communicate with the first processing means 3. The ATM relay apparatus according to claim 1, further comprising a second processing means for managing the switch.

Images