JP2000083062A - Packet transfer method and packet processor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a packet multiplexing method which reduces the delay time of a packet whose delay time request is severe and also highly efficiently utilizes an access line. SOLUTION: An inter-network connecting device 2a divides an IP packet received from a LAN 7a into plural queues on prescribed reference and multiplexes a packet in plural queues where data exist to generate one or more multiple frames. Each multiple frame is sent to a subscriber node 1a by using a logical line differing in accordance with a set (multiplexing mode) of packets in the each multiple frame and the subscriber node 1a identifies the multiplexing mode according to the logical line, divides a partial packet in the multiple frame into plural queues in a procedure being reverse to the procedure of multiplexing, restores the original packet, generates a frame including each packet and defines a logical line to be used according to the IP address of the packet to send it to a relay network 4a.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a packet multiplexing method and a packet multiplexing method suitable for use in a wide area packet switching network including a plurality of interconnected LANs (Local Area Networks) and a packet switching network connecting them. And a packet processing device such as a network connection device or a packet switch.

[0002]

2. Description of the Related Art In a current network technology, a plurality of communication networks are connected to each other, and an IP (Internet) is connected between them.
net Protocol) packet exchange is possible. More recently, packet-switched networks have been
Attempts have also been made to transmit P packets.

[0003] Various types of devices are used to interconnect different communication networks. Hereinafter, these devices are collectively referred to as an internetwork connection device. This device basically accommodates a plurality of data link lines, extracts a packet from a frame received from one of these data link lines, generates another frame including the packet,
It has the function of transmitting to other data link lines. Devices in which the address used for switching the transfer route of the IP packet is the address of the data link layer, the address of the network layer, or the address of a layer higher than the network layer are called a bridge, a router, and a gateway, respectively.

A plurality of LANs provided inside a company or the like are connected to a wide area packet switching network, and a packet switching network capable of exchanging IP packets between these LANs has also been realized. A packet switching network is a communication network composed of a plurality of packet switches. A chunk of data actually exchanged in a packet switch may be called a packet, a frame, a cell, or another name. As is usually the case, here again, regardless of its name,
A switch using the packet switching method is called a packet switch. Packet switching is a term for message exchange, which is a method in which a message is divided into a plurality of short pieces of information and exchanged with a header. Basically, the packet switch determines the destination of the packet from the contents of the packet received from one of the plurality of data link lines, and distributes the packet to at least one output line determined according to the destination.

[0005] Various types of packet switches are used. For example, an ATM switch or a frame relay switch is a typical example. When discussing without distinguishing between an ATM switch and a frame relay switch, A
A cell handled by the TM exchange may be regarded as a frame.
In the following, a packet may be included in a frame. When referring to the packet, it may be simply referred to as a packet.

Alternatively, the frame may be simply called a packet.

[0007] The ATM switch is suitable for handling multimedia information such as voice, data, and video. A chunk of data transferred by the ATM switch is called a cell and has a fixed length. The length of the cell is 53 bytes, of which the header part for identifying the destination is 5 bytes and the information part is 48 bytes. The ATM exchange can not only multiplex and transfer cells belonging to a plurality of logical channels sent from the same terminal via a single physical line, but also
When multiplexing cells transmitted from different terminals, the number of cells from each terminal is dynamically changed according to the transmission speed of those terminals. This can increase the data transmission speed from a terminal that transmits a larger amount of data.

In the ATM switching system, AAL (ATM Adaptation LTE) plays a role of decomposing user information into cells and restoring original user information from cells.
ayer). AAL, a type of AAL
In No. 5, when continuously transmitting a plurality of cells including information belonging to the same IP packet in the same virtual channel, the IP header is included only in the first cell and is not included in subsequent cells. For example, ITU-T
(International Telecommun
ication Union-Telecomuni
station Standardization Sec
tor) Recommendation I. tor) 363.
5 “B-ISDN ATM Adaptation La
Yer Type 5 Specification ",
PP. 1-20 (1996). As a result, the header overhead is reduced because most cells do not include a header portion. However, this method can be used only when one virtual channel does not simultaneously transmit and receive IP packets destined for a plurality of destinations.

[0009] When transmitting voice packets, if the delay of voice transmission is large, it is difficult or difficult to hear in real time. AAL2 employs the following method to reduce the transfer time of voice information. For example, ITU-TRRecommendatio
nI. 363.2 "B-ISDN ATM Adapt
ation Layer Type 2 Specific
ation ", PP. 3-35 (1997). A header for identifying voice information and a destination is stored in a minicell, and a plurality of minicells are stored in one ATM cell.
Will be transferred. That is, a plurality of minicells are multiplexed into one cell. A plurality of (voice) channels are multiplexed in one cell. A header is included in each minicell to identify the channel of each minicell. This header is 3 bytes. At the node receiving this multiplexed cell, minicells sent to different destinations are replaced with ATM cells for each destination.

The frame relay exchange is a packet exchange suitable for high-speed data communication such as connection between LANs. The block of data to be transferred is called a frame, and can transfer a frame of variable length and longer than an ATM cell.
Multiple logical lines assigned a data link connection identifier (DLCI) can be used.

A representative example of a wide area packet switching network in which a plurality of LANs and a frame relay exchange are connected is as follows. A subscriber node is provided corresponding to one or a plurality of LANs of the transmission source. This subscriber node is usually connected to the LAN via an access line dedicated to each LAN and an inter-network connecting device. A frame including an IP packet, for example, an Ethernet frame, is sent from a terminal connected to the transmission source LAN to the transmission source network connection device via the LAN. This device converts the frame into a frame of a format suitable for the packet switching network, and sends the frame to the LAN-compatible subscriber node provided in the network.

This frame is transmitted by the packet switching network.
The data is sent to the destination subscriber node provided corresponding to the destination LAN via the source subscriber node. The destination subscriber node sends this frame to the destination network connecting device connected to the LAN via an access line dedicated to the destination LAN. The destination network connecting device changes the frame into a frame format used in the destination LAN, and transmits the frame to the destination LAN. Voice transmission is also performed using the packet switching network.

JP-A-5-268220 discloses a branch line LAN.
Each branch LAN in the transmission frame of the main LAN connecting
Use multiple timeslots assigned to packets from. Here, the number of assigned time slots is
It is changed according to the expected load of the branch line LAN. The assigned slot is notified from the management device to each branch LAN.

However, basically, a method is adopted in which a data area in a frame is divided into a plurality of areas, and each area is assigned to a specific line.

[0015]

SUMMARY OF THE INVENTION An ATM switching network is used as a packet switching network.
In the case of using a type of converting a voice packet received from the ATM cell into an ATM cell and transmitting the ATM packet to a subscriber node on the transmitting side, the internetworking apparatus divides a plurality of IP packets into a plurality of cells, and A cell for a packet is transmitted on the same physical line when divided. Sending LAN
If a voice packet arrives immediately after a data packet arrives at the transmitting-side internetworking device from the network, the cell containing the voice packet becomes the internetworking device after one of the cells containing the data packet is transferred. Transferred from. Since one cell is 53 bytes, the transmission delay time of the voice packet in the network connection device is short. For example, if the speed of the access line is 1.536 Mbit / s, this transmission delay time can be 276 microseconds or less. This transmission delay time does not depend on the size of the data packet.

However, there is a problem that the ratio (overhead) of the header portion to the entire cell is large, so that the line utilization rate or the effective throughput is reduced from 80% to about 90%. Here, the line utilization efficiency or the effective throughput is the ratio of the time during which information to be transmitted is currently being transmitted to the time used to transmit one data block (for example, a frame or a cell). Specifically, it is a ratio of a portion excluding a header in a frame or a cell.

Further, when the AAL2 method is used in the ATM exchange, a header is provided in each minicell in the cell, so that the overhead is further increased by the amount of the header and the line use efficiency is reduced.

In the AAL5 method of the ATM exchange, only the first cell is provided with a header and the subsequent cells are not provided with a header, so that the overhead can be reduced. However, this method can be used only for a plurality of cells for the same voice packet, and for a plurality of voice packets transmitted from a plurality of users, it is necessary to perform cell transmission in voice packet units. The method is not applicable.

Therefore, in a packet switching network such as an ATM switching network using a packet switch transmitting relatively short cells, data which belongs to a plurality of IP packets such as ALM2 is multiplexed and transmitted in the same cell. A packet multiplexing method suitable for preventing a decrease in line use efficiency is desired.

A type in which a frame relay switching network is used as a packet switching network, and an inter-network connecting device converts a voice packet received from a LAN on the transmission side into a frame for a frame relay exchange and transmits the frame to a subscriber node on the transmission side. Is used, the network connecting device transmits the LA of the transmission source there.
When an Ethernet frame is transferred from N, this frame is converted into a frame for a frame relay exchange. Frames of the frame relay system have a variable length, and long data can be transmitted in one frame. Therefore, the overhead of the header portion of the frame can be reduced.

However, when a plurality of Ethernet frames are transferred from the transmission source LAN to the network connection device, these frames are sequentially converted into frames for the frame relay exchange. Therefore, if any voice packet arrives immediately after a data packet has arrived from the transmitting LAN to the transmitting-side interworking apparatus, the transmission of the frame including the voice packet is performed by transmitting all the frames including the data packet. Does not start until The transmission delay time of a voice packet increases as the data packet increases. Conventionally, frames sent on the frame relay network are LAN
Most of the data is used to connect data, and the frame length is almost 1500 bytes or less due to the limitation of the maximum frame length in Ethernet. However, when an application that transfers a large amount of data is required in the future, it is considered that the number of frames exceeding 1500 bytes increases in order to reduce overhead. However, it is difficult to increase the access line speed from the viewpoint of communication cost. In such a case, if the speed of the access line is 1.536 Mbit / s as described above, for example,
In order to transmit a 4096-byte data packet from the transmitting-side interworking apparatus to the corresponding subscriber node, 21.3 is used.
It takes milliseconds. This time becomes the transmission delay time of the voice packet. A voice packet is usually about 256 bytes in size, but this transmission delay time occurs regardless of the size of the voice packet. This transmission delay time does not satisfy the requirement regarding the delay time of the voice packet.

As described above, in a packet switching network using a packet switch which can transmit a relatively long frame such as a frame relay switching network, there is an advantage that the overhead of the frame is small. When transmitting the packet, the transmission delay time becomes a problem. The same problem arises when transferring voice packets from a subscriber node to an internetwork. In that sense, both the network connection device and the corresponding subscriber node have the same problem. In this specification, when various devices that process packets including these devices are generically referred to, these devices are referred to as packet processing devices.

In the multiplexing method described in Japanese Patent Application Laid-Open No. 5-268220, if data of any line is lost, only invalid data is stored in an area in a frame allocated to that line. No longer included, and the multiplexing method described here is not efficient in line utilization.

Accordingly, an object of the present invention is to reduce the transmission delay time of a specific type of packet such as a voice packet in an internetworking apparatus capable of transmitting a relatively long frame such as a frame for frame relay. To provide a packet transfer method and an inter-network connection device.

A more specific object of the present invention is to provide a packet multiplexing method and an interworking apparatus suitable for reducing the transmission delay time of a particular type of packet such as a voice packet without significantly reducing the line use efficiency. Is to provide.

[0026]

In order to achieve the above object, in a packet transfer method according to the present invention, at least a part of data constituting each of a plurality of packets to be transmitted is added to a header portion of each packet. The belonging data and the data belonging to the data part of the packet are cut out without distinction. A plurality of frames cut out from the plurality of packets are included as transmission data, and a multiplexed frame having header information for transmitting the transmission data to a predetermined receiving apparatus is generated and transmitted. The amount of the data to be cut out from each packet is changed according to the number of the plurality of packets within a range where the length of the transmission data does not exceed a predetermined limit length. More specifically, when there is no more data to be cut out from any one of the plurality of packets, the above-described cut-out and transmission are repeated for a plurality of packets other than the one packet.

[0027] Information for separating each of the plurality of data from the transmission data is transmitted to the receiving side device. The information for separation is desirably included in the multiplexed frame, but may be transmitted in a frame different from the multiplexed frame.

In a more specific mode of the packet transfer method according to the present invention, a plurality of packets to be transmitted are divided into a plurality of predetermined groups according to a predetermined standard. From each of the plurality of groups to which at least one packet belongs, the data belonging to the header portion and the data belonging to the data portion of at least one packet belonging to the group are distinguished from each other. Cut out at least a part of the data constituting. The transmission data includes a plurality of data cut out from the plurality of groups, and does not include data related to other groups other than the plurality of groups among the predetermined plurality of groups, and a predetermined reception of the transmission data. One multiplexed frame including a header portion to be transferred to the device on the side is generated. The multiplex frame is transmitted with information for identifying a group set to which a plurality of packets obtained by cutting out the plurality of data included in the multiplex frame belong.

Each time the receiving apparatus receives a plurality of multiplexed frames transmitted from the transmitting side apparatus and the corresponding information, a plurality of data included in different packets included in each multiplexed frame are received. Are separated based on the information, and each data is stored in correspondence with one of a plurality of predetermined groups corresponding to the plurality of predetermined groups determined by the information. The packet transmitted by the transmitting device is restored from the plurality of data stored in each of the plurality of groups in which the data is stored.

As described above, since the data of a plurality of packets are multiplexed and transmitted in one frame, for example, if the above-mentioned criterion is determined so that the voice packet is classified into another group from the data packet, Even if a large data packet is received immediately before receiving the voice packet, the data of the voice packet is transmitted together with the data of the data packet. Therefore, voice packets do not have to wait until all of the data packets have been transmitted.

Further, since the multiplexed frame does not include identification information corresponding to each data, large extra information due to the multiplexing is not added to the frame, and the line efficiency does not decrease so much.

More specifically, the present invention is implemented as follows.

When there are a plurality of packets belonging to the same group, the plurality of packets to be transmitted are divided into a plurality of packet strings corresponding to the group so that those packets are linked and stored in one packet string. And memorize. In the above-mentioned extraction, at least a part of the data constituting the packet sequence stored corresponding to each group is extracted. Further, data constituting a packet sequence stored corresponding to one of the groups is cut out without discriminating to which of a plurality of packets constituting the packet sequence the data belongs. Transmitting a plurality of packet sequences stored corresponding to a plurality of groups to which at least one packet belongs from the above-mentioned predetermined plurality of groups, from the above-mentioned step of cutting out for a new multiplexed frame to the above-mentioned transmitting step repeat. Each packet sequence is stored in one of a plurality of predetermined transmission queues.

A packet processing device such as an internetwork connecting device or a packet switch according to the present invention is configured to execute the above packet transfer method.

[0035]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, a packet transfer method and packet processing according to the present invention will be described in more detail with reference to some embodiments shown in the drawings. In the following, the same reference numerals represent the same or similar ones. In the second and subsequent embodiments, only the differences from the first embodiment will be mainly described.

Embodiment 1 of the Invention (1) Schematic Configuration of Packet Switching Network FIG. 1 shows an example of a packet switching network to which the present invention is applied. This packet switching network includes LANs 7a, 7b and 7c and relay networks 4a and 4b. Here, it is assumed that the relay networks 4a and 4b are packet switching networks of the frame relay system. However, the present invention is applicable to other packet switching networks.
Each of the LANs 7a, 7b, 7c is connected to the corresponding network connection device 2a, 2b or 2c and the access line 3a,
3b or 3c via the subscriber nodes 1a, 1
b or 1c. Each access line and the network connection device connected thereto are usually provided exclusively for the corresponding LAN. The relay network 4a includes a plurality of relay nodes, for example, 5a and 5a, in addition to the subscriber nodes 1a and 1b.
b, 5c and a plurality of trunk lines such as 6a, 6b, 6c, 6
d, 6e, and 6f.

The relay network 4b includes, in addition to the subscriber node 1c,
It includes a plurality of relay nodes such as 5d and 5e and a plurality of relay lines such as 6g, 6h and 6i. The inter-network connecting device 2d is further connected to the relay network 4a by a relay line 6f to the relay node 5b, and the inter-network connecting device 2e is further connected to the relay network 4b.
Are connected to a relay node 5d by a relay line 6h, and the relay networks 4a and 4b are connected by these inter-network connecting devices 2d and 2e and a relay line 6g. Subscriber node 1
a, etc., and the relay node 5a, etc. are constituted by packet switches. The network connection device 2a or the like is formed of a device called a router or a gateway (in a narrow sense). They are,
It is suitable for connecting a plurality of networks to form a wide area communication network. However, the present invention can also be applied to other network connecting devices other than these specific network connecting devices.

In the present embodiment, the internetwork connecting device 2a or the like transmits a plurality of partial data obtained by dividing a plurality of IP packets included in a plurality of Ethernet frames supplied from the LAN 7a or the like connected thereto. Generate one or more multiplex frames including the corresponding subscriber node 1
a. Subscriber node 1a
Decomposes the data portion of each multiplexed frame, restores each IP packet from a plurality of partial data included therein, reconstructs a frame relay frame including each IP packet, (Not shown)
Is configured to be forwarded to.

Further, the subscriber node 1a is connected to the LAN 7a
Is configured to multiplex a plurality of frame relay type frames to be transmitted to the network connection device 2a,
The network connection device 2a is configured to disassemble the network frame and generate an Ethernet frame including a plurality of IP packets. The same applies to other network connection devices and subscriber nodes.

(2) Outline of Packet Transfer Operation In this embodiment, as shown in FIG. 2, a data link is provided on an access line 3a between a transmitting-side interworking apparatus 2a and a corresponding subscriber node 1a. Multiple logical lines 9 using
A, 9B,... 9C are set in advance, and virtual multiplexing circuits 20A, 20
B or 20C and the virtual separation circuits 10A and 10A
B, or 10C can be considered equivalent to being installed. The virtual multiplexing circuits 20A, 20B, and 20C located at the transmitting end point read data from the transmission queue by different multiplexing methods, multiplex the data into frames, and virtual multiplexing circuits 10A and 10A located at the opposite receiving end points. 10B or 10C performs the inverse transformation.
The transmission-side network connecting device 2a selects a multiplexing circuit that performs optimal multiplexing from the states of the plurality of transmission queues, multiplexes data into a single frame, and transmits the data to one of the logical lines. Therefore, a normal data link can be used for the part of the logical line. The above applies to a case where a frame is transmitted from the subscriber node 1a to the network connection device 2a, and also to a combination of another network connection device and a subscriber node.

(3) Outline of Packet Multiplexing As shown in FIG. 3A, the header part of the Ethernet frame 40 received from the LAN 7a by the network connection device 2a includes a preamble 40A, a destination MAC address 40B, a source MAC address 40C, The frame 40 includes data 41 and a frame check sequence (FCS) 40E.

This data 41 is composed of an IP packet.
The IP packet is composed of an IP header 41A and a data part 41B. The IP header 41A contains the IP address of the destination of the packet.
Contains the address. The data section 41 is composed of data of a layer higher than the IP layer, and specifically, is composed of a header of the TCP layer or a header 42A of the UDP layer and a data section 42B.

The internetwork connection device 2a extracts the IP packet 41 from the Ethernet frame 40, encapsulates it with PPP (point-to-point protocol), and
3 is generated. At this time, the header part in the Ethernet frame 40 is discarded. The purpose of using PPP is to enable the processing of this IP packet by a multi-protocol, and is not necessary for the multiplexing method of the present invention.

As shown in FIG. 3B, the PPP frame 43 has a flag sequence 43 indicating the head of the frame.
A, an address section 43B, a control section 43C, a protocol section 43D, a data section 43E, a frame check sequence 43F, and a flag sequence 43G indicating the end of the frame. The data part 43E includes the IP packet 41
Is stored as is.

At the time of this PPP encapsulation, an octet stuffing process is performed. The octet stuffing process includes the first flag sequence 43A and the last flag sequence 4A of the generated PPP frame.
This is a process of changing the bit sequence in the generated frame so that the same bit sequence as these flag sequences is not included during 3G. With this processing, the subscriber node 1a can later determine the start and end of the PPP frame using the flag sequence.

A plurality of transmission queues 1 are provided in the network connection device 2a.
(35A), transmission queue 2 (35B), transmission queue 3
(35C), and a transmission queue 4 (35D) are provided.
This frame is classified based on the P header 41A and the TCP or UDP header 42A. That is, a transmission queue for storing the frame is selected, and this frame is written to the selected transmission queue.

In the present embodiment, a plurality of frames to be transmitted are classified into a plurality of groups. Each transmission queue is an area for linking a plurality of frames to be transmitted belonging to the same group and storing them as a frame sequence. The same applies to the reception queue described later. The frame classification criteria, that is, more specifically, here, the criteria for selecting the transmission queue for storing the frame, can be appropriately determined by the system administrator. With this selection criterion, a PPP frame including a specific type of IP packet can be assigned to a specific transmission queue. For example, I
Since a P packet usually includes a specific transmission port number in a TCP header or a UDP header, a PPP frame including a transmission port number corresponding to a voice packet is assigned to a specific transmission queue, for example, transmission queue 1, and other than that. An IP packet, for example, a data packet can be equally allocated to the other transmission queues 2, 3, and 4. In this assignment, voice packets in the transmission queue 1 can be transmitted with priority. It is also effective to transmit voice packets together with other IP packets stored in the transmission queues 2, 3, and 4. Also, a PPP frame having a specific destination IP address can be stored in a specific transmission queue. These selection rules can be combined. As described above, in the present specification, not only packets having different data but also packets having different IP addresses, such as voice packets or data packets, are considered to be different types of packets. Generally, if the packet is classified into a plurality of sections according to a predetermined criterion, it is considered that those packets have different types.

When the PPP frame is written in at least one transmission queue, the network connection device 2a starts the transmission operation. In this transmission, when there are a plurality of transmission queues storing PPP frames, the partial data of the PPP frames in those transmission queues is multiplexed into one frame relay frame 46 and transmitted. The header part of the frame 46 includes a head flag sequence 46A, an address part 46B, a control part 46C, and a network layer protocol identifier (Network Layer Pr).
otocol Identifier-NLPID) 4
6D, the frame further includes a data part 46E, a frame check sequence part 46F, and a tail flag sequence 46G. Control section 46C,
The NLPID unit 46D uses an IETF (Internet) to carry multi-protocol packets by frame relay.
RFC (Request for Comments) 1 of t Engineering Task Force)
490. Address part 4
6B is based on ITU-T standard Q.6. 922, and Q.922. 922 address part. DLC of logical line
I is included in this address part.

The data section 46E includes a plurality of transmission queues 1
And a plurality of partial data obtained by dividing a plurality of PPP frames within the range from. When the PPP frame is divided, the PPP frame is divided without distinguishing the flag sequence, the header part, the data part, and the frame check sequence. The combination of the partial data is performed only for the transmission queue in which the PPP frame exists. For example, assuming that at least one PPP frame exists in the transmission queues 1, 2, and 4, and no PPP frame exists in the transmission queue 3, the partial data in the transmission queues 1, 2, and 4 are stored in the data section one by one. 46E, and furthermore, the transmission queue 1,
Subsequent partial data in 2 and 4 are stored one by one, and so on. The formation of the data section 46E ends when the PPP frame in any of the transmission queues disappears, the length of the frame 46 reaches a predetermined maximum length, or another predetermined condition is satisfied. The octet stuffing process is performed on the multiplex frame. The octet stuffing process changes the bit sequence of the generated frame so that the same bit sequence as these flag sequences is not included between the head flag sequence 46A and the tail flag sequence 46G of the generated multiplex frame.

When the length of the multiplex frame 46 reaches the predetermined maximum length, a new multiplex frame is generated for the PPP frames in the transmission queues 1, 2, and 4. Alternatively, when there are no more PPP frames in any of the transmission queues, new multiplex frames are generated for the PPP frames in the other transmission queues. As exemplified above, when a PPP frame including a voice packet is stored in the transmission queue 1 and a PPP frame including other packets is equally stored in the other transmission queues 2, 3, and 4, the voice packet is transmitted. The partial data of the included PPP frame can be multiplexed together with the partial data of another PPP frame in the same frame. As a result, even when a voice packet is received by an internetworking device after another data packet, transmission of the voice packet can be started without waiting for all of the data packets to be transmitted. Thus, it is possible to reduce the transmission delay time of the voice packet in question.

The length of the partial data cut out from each transmission queue can be changed for each transmission queue. Specifically, a threshold value of the number of frames is set in advance corresponding to each transmission queue. The internetwork connection device 2a determines whether the transmission queue stores a PPP frame in a transmission queue and determines whether the number of frames stored in the transmission queue exceeds a threshold value defined for the transmission queue. The length of the partial data read from the transmission queue can be changed based on the accompanying information such as whether or not the data is included.

As will be described later, the device on the receiving side combines the PPP frame included in the multiplexed frame,
Therefore, different partial data in the multiplexed frame is cut out using information for identifying a combination of IP packets. Therefore, this identification is information for separating a plurality of partial data in the multiplexed frame. This combination represents a multiplexing mode. This multiplexing mode is different for each multiplex frame.

In this embodiment, the combination can be determined by the combination of the group to which the packet from which the data has been read belongs, that is, the combination of the transmission queue from which the data has been read. In the present embodiment, the information notifying the multiplexing mode of a certain frame includes the identification number DL of the logical line corresponding to the multiplexing mode of the multiplexed frame.
CI is used.

That is, a plurality of logical lines respectively corresponding to one multiplexing mode are newly defined for transmitting a multiplex frame between the network connection device 2a and the subscriber node 1a. Frames multiplexed in different multiplexing modes are transferred using different logical lines. As will be described later, the address section 46B includes the DLCI of the logical line used when transmitting the frame 46. Frame 46
Is generated and stored in the address section 46B using the DLCI corresponding to the multiplexing mode used at that time. The DLCI included in the address part is used as information for identifying the used multiplexing mode. Therefore, in this embodiment, there is no need to send this information separately.

(4) Outline of Packet Restoration Operation As shown in FIG. 4A, a plurality of reception queues 1 (65A), 2 (65B), 3 (65C), 3 A queue 4 (65D) is provided. The number of these queues is the same as the transmission queues 35A to 35D described above.

Upon receiving the multiplex frame 46, the subscriber node 1a distributes a plurality of partial data included in the data part 46E of the frame to these reception queues. This distribution corresponds to D in the address portion 46B of this frame 46.
This is performed based on LCI.

That is, each partial data is stored in the reception queue corresponding to the transmission queue from which each partial data has been read. Therefore, the PPP frames held in the transmission queues 1 to 4 are stored in the reception queues 1 to 4.

As shown in FIG. 4B, the subscriber node 1a
Performs PPP decapsulation on the stored partial data when the partial data is stored in any of the reception queues.

That is, a flag sequence is found from the information written in the reception queue. When a flag sequence is found, a PPP frame 48 delimited by the flag sequence is extracted, and the PPP frame 48 is extracted.
Out of the IP packet 48E. Further, the subscriber node 1a re-encapsulates the IP packet 48E into a frame 49 of a frame relay and sends it to the relay network 4a.

The header part of the frame 49 includes a flag sequence 49A, an address part 49B, a control part 4
9C and an NLPID unit 49D. This frame includes a data section 48E, a frame check sequence 49F,
A flag sequence 49G is included, and an IP packet 48E is stored in the data section 48E. IP packet 48E
Includes an IP header 48H, a TCP / UDP header 48I and a data part 48J, and includes
The LCI connected to the destination terminal (not shown) is connected to the LCI.
N of the logical line leading to the destination network connection device corresponding to N
LCI is stored. The destination terminal is indicated by the IP address in the IP header 48H.

The subscriber node transmits the frame relay frame received from another relay node to the internetwork
Also, when sending the frame to a, a plurality of frames can be multiplexed in the same manner as performed by the network connection device 2a. Upon receiving the multiplexed frame, the internetwork connecting device 2a can perform the same process as the restoration process of the packet from the multiplexed frame performed by the subscriber node 1a.

(5) Details of Network Connection Device (5A) Device Configuration As shown in FIG. 5, the network connection device 2a is provided with a program-controlled processor 27A and a randomly accessible memory (RAM) 28A. ing. Processor 27
The program for A is held in the memory 28A or a ROM (not shown). In the network connection device 2a,
A physical layer reception processing unit A (21A) for receiving an Ethernet frame from the LAN 7a;
a, a physical layer transmission processing unit A (26A) for transmitting an Ethernet frame to the access line 3a; a physical layer transmission processing unit B (23A) for transmitting a frame relay frame to the access line 3a; Physical layer reception processing unit B (24A) for receiving the frame of the frame check sequence (FC)
S) a calculation circuit. Other network connecting devices have the same structure.

When generating a frame relay type frame including an IP packet in a received Ethernet frame, the conventional network connection apparatus should transmit the frame to be generated by looking at the IP address included in the packet. The destination, for example, the network connection device was determined. The DLCI of a frame to be generated is defined so as to use a logical line defined between the destination network connecting apparatus and the self. However, in the present embodiment, the network connection device 2a is configured to generate a multiplex frame from a plurality of Ethernet frames.

FIG. 6 shows various areas provided in the memory 28A and used for multiplex frame generation. The transmission queues 1 to 4 (35A to 35D) already described are also provided in the memory 28A. Although the number of the transmission queues is four as an example here, this number can be changed as appropriate. The size of each transmission queue may be determined depending on the arrival process of the input packet. Increase the size of the transmit queue where many packets arrive.
For example, it is set to several tens of kilobytes to several megabytes. However, in the case of a transmission queue having a large amount of data to be read at the time of generating a frame, or a transmission queue having a high priority, which will be described later, the memory size of the transmission queue can be smaller than that of other transmission queues. Similarly, FIG.
Indicates data stored in the internal register group of the processor 27A.

(5B) Ethernet Frame Reception Operation Upon receiving an Ethernet frame from the LAN 7a, the physical layer reception processing unit A (21A) issues a reception interrupt to the processor 27A. Processor 27A executes the reception process shown in FIG. 10 in response to this interrupt.

The processor 27A sends the Ethernet frame 40 (FIG. 3) from the physical layer reception processing unit A (21A).
A) is read (100), and it is confirmed that the type 40D of the read frame header is 0x-08-00 defined for the Ethernet frame, and that the encapsulated data is an IP packet ( 10
1). Here, 0x indicates that the data to which it is attached is data in hexadecimal notation.

The processor 27A classifies the IP packet according to a predetermined standard, and classifies the IP packet according to its type.
The transmission queue in which the P packet is to be stored is determined as follows. The IP header 41A (FIG. 6) of the IP packet includes a source IP address, a destination IP address, a packet length, and other information. Also, TCP or UD
The P header 41B (FIG. 6) includes a transmission port number (also called a source port number), a reception port number (also called a destination port number), and other information. Source IP
Both the address and the destination IP address are divided into a network address part for specifying an IP subnet and a host address part for specifying a host in the IP subnet.
The destination IP address is used to identify a destination terminal of an IP packet. A specific queue can be assigned to IP packets having a specific destination IP address for a specific predetermined user.

The transmission port number and the reception port number are
This is a number for identifying an application program that transmits or receives an IP packet. The application program on the transmission side uses a port number determined by itself, but the destination application program usually uses a specific port number to identify a higher-level protocol. For example, an application program that executes a specific process such as an application program that handles voice packets has a well-known port number (well known port num).
ber) using a specific port number. Therefore, it is possible to determine whether or not the IP packet is a packet containing a specific type of data, for example, voice, based on the reception port number included in the IP packet.

As shown in FIG. 8, the queue correspondence table 3
1 holds the number of a transmission queue in which an IP packet having a different set of a destination IP address, a mask, and a reception port number is to be stored. Here, a specific transmission queue to be assigned to a combination of a specific destination IP address and a specific reception port number is defined, but this assignment can be freely set. In addition,
The mask is for masking a portion of the IP address that is not used for determining the queue number. The IP address is
Network address part and I that specify IP subnet
It is divided into a host address part that specifies a host in the P subnet. The mask is used to extract the network address part from the destination IP address. Here, a destination address having a network address of a specific network, for example, a LAN, is treated as a specific IP packet, and a specific transmission queue is assigned to the IP address.

For example, an I-port having a specific reception port number
For a P packet, a specific transmission queue, for example, transmission queue 1, is allocated without depending on its destination IP address. Can be equally assigned. In this case, if a reception port number for voice processing is used as the specific reception port number, this assignment is effective when transmitting only voice packets with priority. In addition, an IP address having a specific destination IP address
A specific transmission queue can be assigned to the P packet without depending on the reception port number.

The processor 27A fetches the destination IP address and transmission port number in the received IP packet (102), and uses the fetched destination IP address and transmission port number as a key in the queue correspondence table 31 (FIG. 6). ) To search for a transmission queue in which the IP packet is to be stored (103). That is, the processor 27A compares the extracted destination IP address with the destination IP address in the queue correspondence table 31, and further,
The extracted transmission port number is compared with the transmission port number in the queue correspondence table. If the destination IP address and the transmission port number also match, the queue number in that row indicates the queue that stores the received packet. If destination I
If either the P address or the transmission port number does not match, the next line is searched until it matches. If they do not match up to the end, they are stored in a predetermined queue. In this way, a queue for storing the received packet is determined.
The network administrator configures the queue correspondence table,
Change according to addition or deletion of contractor, increase or decrease of traffic, etc. In order to screen service contractors,
A transmission queue can be assigned to a pair of a source IP address and a destination IP address.

Next, the processor 27A checks whether or not the free space in the transmission queue determined as described above is smaller than the IP packet (110). If the empty area is smaller than the packet length, the received IP packet is discarded (120), and the process ends. However, as the size of the IP packet, the length when the IP packet is encapsulated in a PPP frame is used. For this check, the queue management table 32 (FIG. 6)
Use

As shown in FIG. 9, the queue management table 3
Reference numeral 2 includes a start address and an end address, a start pointer, an end pointer, the number of bytes, a threshold value, a waiting data transmission interrupt flag, and a transmission interrupt flag exceeding the threshold value, corresponding to each transmission queue. The start address and the end address indicate the start address and the end address of the area allocated to the corresponding transmission queue in the memory 28A. This transmission queue area is used cyclically,
The start pointer and the end pointer indicate the start address and the end address of the area where the IP packet is actually stored. The number of bytes indicates the total number of bytes of all IP packets stored in the transmission queue. The threshold value, the waiting data transmission interrupt flag, and the transmission interrupt flag exceeding the threshold value will be described later. The processor 27A calculates the size of the free area in the queue management table 32 from the start address, the end address, and the number of bytes.

If the free area is larger than the packet length, the processor 27A encapsulates the received IP packet in PPP as follows and stores it. Encapsulation by PPP is because the resulting PPP frame is suitable for processing with various protocols.

First, the processor 27A checks whether or not the data of the IP packet previously stored in the transmission queue remains in the transmission queue by checking whether or not the total number of bytes of the transmission queue is 0 (130). ). if,
When there is no waiting data in the transmission queue, the bit sequence 0x-7E is stored in the reception frame work area 1 (36A (FIG. 6)) in the memory 28A as the flag sequence 43A indicating the head of the PPP frame (131). As described later, a flag sequence 44G (FIG. 6) is added to the end of each PPP frame. If the next PPP frame is stored after that frame, this subsequent PPP frame
The flag sequence at the beginning of the frame can be omitted. For this reason, if there is waiting data in the transmission queue, the processing 1
31 is not performed.

Next, the processor 27A transmits the IP in PPP.
According to the packet encapsulation method, the address string 43B, the control section 43C, and the protocol section 43D are stored in the received frame work area 1 (36A) as bit strings 0, respectively.
x-FF, 0x-03, 0x-21 are added in order (1
40). 0x-FF of the address section 43B indicates all terminals by a global address. The control unit 43C
The control unit 43C is provided for adjusting the PPP frame to the HDLC frame format.
0x-03 indicates an unnumbered frame. The protocol section 43D is defined for making the PPP frame compatible with the multiprotocol, and indicates the protocol of the data section 43E. The protocol unit 43D is usually 2 bytes, and in the case of an IP packet, it becomes a bit string 0x-00-21. However, in order to reduce the overhead, it is also possible to negotiate between transmission and reception and compress it to 1 byte. . At that time, the protocol part for the IP packet is a bit string 0x-21. In the present embodiment, this compressed protocol section is used.

The received IP packet 43E is stored after those bit strings in the received frame work area 1 (1
50). Thereafter, the processor 27A calculates a frame check sequence (FCS) 43F from the address section 43B to the end of the IP packet 43E using the FCS calculation circuit 29A (FIG. 5), and receives the received frame work area 1F.
(160) is added to (36A). Even if a part of the generated PPP frame is discarded during transfer and a part of the data in the PPP frame is lost, the discard of the data can be detected by checking the frame check sequence 43F. Therefore, this frame check sequence is necessary regardless of whether or not to encapsulate by PPP. Note that a packet whose data part is partially discarded is discarded.

Processor 27A performs octet stuffing processing 170 on the data in received frame work area 1. Octet stuffing processing 170
Is a flag sequence 43 at the head of the generated frame.
This is a process of changing the internal bit sequence of the generated frame so that the same bit sequence as the flag sequence 0x-7E is not included between the flag sequence 43A and the flag sequence 43G at the end. Can be detected correctly. Therefore, this processing is necessary regardless of whether or not to perform encapsulation based on PPP. This process is based on IETF's RFC1662 "PPP in HDLC
-Like framing "in the following manner. In this case, the received frame work area 1
(36A) is used as an input area, and the transmission queue is used as an output area.

Referring to FIG. 11, processor 27A reads one byte of data in the input area (200). Next, the read data is checked (210, 220, 23).
0), if the read data is 0x-7E, the data is changed to 2-byte data 0x-7D-5E (215). When the read data is 0x-7D,
It is changed to 0x-7D-03 (225), and in the case of 0x-03, it is changed to 0x-7d-23 (235). Data 0
The conversion of x-03 is performed when the control code 0x-03 of the asynchronous line is used in the negotiation. In addition,
If the option was negotiated to use the control code of the asynchronous line, RFC 162 if the control code appeared during the flag sequence.
2 is converted.

The processor 27A stores the converted data in the output area (240), checks whether or not the data in the input area has been completed (250). Returns to the process 200 to continue the process. Thus, the data obtained as a result of the octet stuffing process 170 is written to the transmission queue.

Returning to FIG. 10, the processor 27A adds a flag sequence indicating the end of the frame to the end of the data in the transmission queue (180). This flag sequence is used when the PPP frame is later multiplexed, so that the P1 in the multiplex frame received by the subscriber node 1a.
Used to identify PP packet boundaries. Therefore, this flag sequence is necessary regardless of whether or not encapsulation by PPP is performed.

Thus, a PPP frame is generated and stored in a specific transmission queue. Finally, processor 27A
Updates the queue state for that transmit queue (1
90). The processing is performed as follows.

The queue status register 51 (FIG. 7) provided in the register area inside the processor 27A has a 2-bit area assigned to each transmission queue, and within that area, a 2-bit transmission queue status code. Is stored. The queue status register 51 is reset to 0 in the initial state, and the queue status code for each transmission queue has the value 0
It is set to 0. This status code indicates that the total number of bytes of the IP packet held in each transmission queue is 0. If there is an IP packet in any of the transmission queues,
The transmission queue status code for that transmission queue is set to a value greater than 00, as described below.

Referring to FIG. 12, in the queue status update processing 190, first, the processor 27A sets the queue status code of all the transmission queues of the queue status register 51 to 0.
It is checked whether it is 0 (300). The queue status code of 00 for all transmission queues means that no IP packets are held in all transmission queues and no multiplex frame transmission is performed until a PPP frame is generated first. Show. Therefore, when the queue status codes for all the transmission queues are 00, the transmission restart flag is set to True in order to request transmission of the previously generated PPP frame (30).
1). Otherwise, the transmission start flag is set to False (302).

Next, the start address and end address of the currently used area of the transmission queue storing the previously generated PPP frame are written in the entry of the head pointer and the entry of the end pointer of the queue in the queue management table 32, respectively ( 310). Further, the current number of bytes of the queue is calculated and written to the entry of the number of bytes of the queue in the queue management table 32 (311).

Next, the processor 27A performs the process of setting the status code corresponding to the queue in the queue status register 51 and the process of the transmission process end interrupt flag as follows. A threshold value is stored in the queue management table 32 in advance for each transmission queue. This threshold is a predetermined threshold for the total number of bytes of the IP frame in the transmission queue. Processor 27A
Compares the threshold value with the number of bytes for the transmission queue (320). When the total number of bytes is 1 or more and does not exceed the threshold value stored in the queue management table 32 for the transmission queue, if the queue status code so far is 00, the queue status code is 01. Is performed (330).

The queue management table 32 (FIG. 9) holds a waiting data transmission interrupt flag for each transmission queue in advance. The waiting data transmission interrupt flag stored for each transmission queue is set when the number of bytes of the PPP frame held by the transmission queue changes from 0 to 1 or more, that is, when the transmission queue becomes empty from the PPP state. This indicates whether or not the processor 27A should issue an interrupt requesting the end of transmission of a frame being executed when the state changes to a state in which a frame is held. Process 330
Is executed, the waiting data transmission interrupt flag corresponding to the queue in the queue management table 32 is set to True.
It is determined whether or not (340). This flag is True
In this case, the transmission processing end interrupt flag is set to True (370). In the case of False, the transmission process end interrupt flag is not changed to False. If the queue status code of the transmission queue is already 01 before executing the process 330, the process 340,
370 is skipped.

As a result of the determination in the process 320, if the number of bytes of the PPP frame held in the transmission queue is equal to or smaller than the threshold value stored for the transmission queue, the queue status code of the transmission queue is If it is not already 11, the queue status code is set to 11 (350). Processing 3
In 60, when it is determined that the waiting data transmission interrupt flag corresponding to the queue in the queue management table 32 is True, or when it is determined in step 361 that the transmission interrupt flag exceeding the threshold is True,
The transmission processing end interrupt flag is set to True (37
0) In other cases, the transmission process end interrupt flag is not changed to False. If the queue status code of the transmission queue is already before executing the process 350, the process 360 is skipped, and if the queue status code of the transmission queue is already 11 before executing the process 350, , Processing 360, 361,
370 is skipped.

As a result of the process 340 or 360, when the transmission process end interrupt flag is set, the interrupt flag requests that a new frame including a new PPP frame stored in the transmission queue be immediately transmitted. Will do. Wait data transmission interrupt flag is T
If it is not true, the transmission end interrupt flag is not set even if the transmission queue changes from a free state to a state holding IP packets. Processor 27A continues to generate the frame currently being generated. When a new frame is generated by the processor 27A after the predetermined termination condition is satisfied and the generation of the frame is completed, the IP packet in the transmission queue converted from the empty state to the state in which the IP packet is held is updated. Built into a simple frame.

As a result of the process 361, when the transmission process end interrupt flag is set, the interrupt flag is used to create a frame including the PPP frame stored in the transmission queue even if the frame is currently being created. It will abort and request that a new frame containing more bytes of the PPP frame in the transmission queue be immediately transmitted.

The processor 27A performs the processing 340, 361
Alternatively, after 370, the new queue status code determined in the process 330 or 350 is stored in the area of the queue status register 51 corresponding to the queue (380). Further, the transmission resumption flag is checked (390).
In the case of True, a transmission restart interrupt is issued and a request is made to restart reading of the PPP frame from the transmission queue (391).

Thus, the updating of the queue state is completed, and the reception processing of the Ethernet frame is also completed.

(5C) Multiplexed Frame Transmission Operation As a result of the queue state update processing, if the transmission restart interrupt flag is set in processing 301 (FIG. 12), the processor 2
7A receives a transmission restart interrupt. The processor 27A starts the frame transmission processing shown in FIG. 13 in response to this interrupt.

The partial data obtained by dividing the PPP frame held in the plurality of transmission queues is multiplexed in the same frame. At this time, it is necessary to determine a combination of the transmission queue for reading the partial data and the length of the partial data to be read. This combination is called a multiplexing mode. Various types of multiplexing modes can be defined. Each multiplexing mode will represent a different combination of the IP packet and its length contained in the multiplexed frame.

The basis of setting the multiplexing mode is that data is not read from the transmission queue having no waiting data. As for reading from the transmission queue having the waiting data, the amount of data read from each transmission queue can be the same between the transmission queues. However, the data amount can be changed. For example, a priority may be assigned to each transmission queue in advance, and the amount of reading from a transmission queue with a higher priority may be increased. Further, the amount of data read from the transmission queue having a large amount of data waiting to be transmitted can be increased. Alternatively, the priority and the amount of data waiting to be transmitted can be combined. That is, it is possible to increase the amount of reading from the transmission queue with the higher priority, and to increase the amount of reading from the transmission queue with more data waiting to be transmitted between transmission queues of the same priority. The amount of data read from each transmission queue can be changed by other methods. For example, weight each transmit queue,
The amount of data read from each transmission queue can be changed by weight.

The amount of data read from the transmission queue in the multiplex mode is determined by the policy of the network operator using the present invention. For example, when providing a service for allocating the bandwidth equally to each user, the readout amount is set uniformly. In addition, when providing a service in which the available bandwidth varies according to the usage fee, the readout amount is set to a different value for each user. Further, the readout amount may be set to a different value for each application, not for each user.

As described above, in the queue management table 32 (FIG. 9), the threshold value and the transmission interruption flag exceeding the threshold value are set for each transmission queue. If the amount of PPP frames held in the transmission queue whose threshold exceeded transmission interrupt flag is set to True exceeds the threshold set for the transmission queue, the weight of the transmission queue is increased, The amount of data read from the transmission queue is increased.

As shown in FIG. 6, the area 33 in the memory 28A includes a plurality of pieces of multiplexing mode information 1, 2, 3,.
M (33A, 33B, 33C,..., 33D). A multiplexing mode information area search table 30 for searching for such multiplexing mode information is also provided in the memory 28A.
Is stored in In this table 30, various values that can be taken by the queue status register 51 and the transmission queues 1 to 4 are stored.
For each of the combinations with various possible threshold over-transmission interrupt states that can be used, the start address of the area storing the multiplexing mode information to be used is stored.

Each multiplexing mode information stores a set of the number of a transmission queue to be used and the number of bytes read from each transmission queue. When one of the transmission queues is not used, an invalid value is written in the queue number and the number of read bytes. Each multiplexing mode information may further include the number of assigned queues indicating the total number of transmission queues used in the multiplexing mode, the maximum information length indicating the maximum length that the generated multiplexed frame can take, and the number of frames to be generated. The maximum number of invalid bytes indicating the maximum length of invalid data that can be generated and the address portion used for the generated multiplex frame are stored.

[0100] This address portion includes a Q.264 containing a DLCI assigned to a logical line transmitting a frame generated using the multiplexing mode information. 322 holds information constituting an address portion. This address part is used as the address part of the multiplex frame. The DLCI of the address part of the multiplexing mode information is uniquely defined with respect to the multiplexing mode information, and the receiving apparatus (the subscriber node 1a in this example) determines which DLCI based on this DLCI. It is possible to know whether the multiplexing mode information has been used. The receiving device is used to decompose the received multiplexed frame in a manner dependent on the multiplexing mode specified by the multiplexing mode information.

Since the logical line used when the multiplex frame is transmitted is defined between the network connection device 2a and the subscriber node 1a, the destination of the multiplex frame is transmitted to the subscriber node 1a. Become. Therefore, the logical line in which the multiplex frame is used is determined in advance regardless of the destination of the IP packet to which the plurality of partial data stored in the multiplex frame belongs.

FIG. 14 shows some examples of multiplex mode information. The first multiplexing mode information is used when there is data in all transmission queues. Four transmission queues are used for multiplexing, the maximum information length is 40 bytes,
Maximum invalid byte length is 4 bytes, address part is 0x-80
−01, the number of bytes read from queue 1 is 4 bytes,
4 bytes read from queue 2, queue 3
The number of bytes read from the queue 4 is 4 bytes, and the number of bytes read from the queue 4 is 4 bytes. If the address part is 0x-80-01, the DLCI is 10,000
00000 (binary number). Q. 32
The C / R, E / A of the first byte of the second address part, FECN, BECN, and DE of the second byte are set to 0, and the E /
Since / A is set to 1, the address portion including this DLCI is 0x-80 as a 2-byte address portion.
It becomes -01.

The second multiplexing mode information includes the transmission queue 4
Used when there is data other than. The third multiplexing mode information is used when there is data other than the transmission queue 3. The fourth multiplexing mode information is used when there is data other than the transmission queue 2. The fifth multiplexing mode information is used when there is data only in the transmission queue 1. In any of the multiplexing mode information illustrated here, the length of data read is constant regardless of the transmission queue.

The maximum information length of 40 bytes is merely an example. To reduce the frame overhead,
The larger the maximum information length is, the better, and in order to switch the multiplexing mode in response to the change of the state of the transmission queue, the shorter the maximum information length is, the better the trade-off may be.
The same applies to the number of read bytes 4.

The maximum number of invalid bytes is used for the following purposes. As described later, the transmission queue to be read may become empty during the creation of the multiplex frame. In this case, the creation of the frame is not immediately terminated, but invalid data is included in the frame instead of the read data of the queue, and the generation of the frame is continued. As a result, the generation of the frame is continued using the valid data held in another transmission queue to be read. This reduces the processing overhead associated with interrupting frame generation and resuming frame generation in a new multiplexing mode. But with this method,
Since the overhead required to transmit the same valid information increases, in order to minimize the overhead, if invalid information is transmitted to some extent, the frame generation is stopped and the multiplexing mode is changed. The maximum number of invalid bytes is used to decide to stop frame generation. This maximum number of invalid bytes is also determined appropriately according to the system to be used.

The setting of the maximum information length and the maximum invalid byte length greatly affects the transmission waiting delay time and overhead.
As described above, the overhead can be reduced by setting the maximum information length to be large, but the transmission delay time associated with the multiplexing mode switching increases. Therefore, these values need to be determined based on the line speed, the packet length distribution of each arriving IP packet transmission queue, the arrival interval distribution, and the like. For example, when the line speed is 1.536 Mbit / s, the maximum information length is set to 1000 when the transmission waiting time accompanying the multiplex mode switching is set to 5 ms or less.
It must be less than bytes. However, this is not the case when the waiting data transmission interrupt flag is used. When the overhead is set to 10% or less, it is necessary to set the maximum information length to 100 bytes or more.

Referring to FIG. 13, the processor 27A
First searches the multiplexing mode information area search table 30 (FIG. 6), reads the start address of the area storing the multiplexing mode information to be used, and reads the multiplexing mode start address register which is an internal register in the processor 27A. 5
2 (FIG. 7) (400). The multiplexing mode to be used is selected based on the queue status of the transmission queues 1 to 4 and the transmission interruption status exceeding the threshold for each transmission queue. Here, the transmission interruption state exceeding the threshold is defined as the transmission interruption flag exceeding the threshold is set to True for the transmission queue, and the amount of data in the transmission queue is set for the transmission queue. Indicates whether the threshold value has been exceeded. Therefore, in the above search, the value of the queue status register 51 and the transmission interrupt status exceeding the threshold for the transmission queues 1 to 4 are used.

Note that the transmission interruption state exceeding the threshold value can be considered as an incidental state of the transmission queue. Other types of incidental queue states may be used. Therefore, in the present embodiment, it is said that the multiplexing mode is selected based on the status of the transmission queue and the status of the associated queue. Of course, it is needless to say that the threshold or the transmission interruption flag exceeding the threshold may not be used. Further, in the present embodiment, the queue state is not determined based on the transmission interruption state exceeding the threshold, but such a state may be adopted. When the queue state is determined in this way, the multiplexing mode can be selected according to the queue state.

The processor 27A reads the number of allocated queues, the maximum information length, and the maximum number of invalid bytes from the multiplexing mode information indicated by the multiplexing mode start address register 52, and reads them out of the internal register 5 in the processor 27A.
3, 55 and 56 (FIG. 7) (420). The internal registers 57 and 58 in the processor 27A that hold the information length of the generated frame and the number of invalid bytes indicating the length of invalid data included in the frame are set to 00 (430).

Next, the processor 27A generates a multiplex frame 46 (FIG. 3B) as follows. The address part in the multiplexing mode information indicated by the multiplexing mode head address register 52 is read, and the DLCI contained therein is read.
And generates the address portion 46B (FIG. 3B) of the frame of the relay frame to be generated and stores it in the transmission frame creation area 1 (37A (FIG. 6) (435). As shown in FIG. Address part 46 of the frame
B is an ITU-T standard Q.B. 922, including the upper 8 bits and the lower 2 bits of the DLCI in separate bit positions. The area for this address part 46B is set to the transmission frame creation area 1 (37
A (FIG. 6)). The 1-byte DLCI read from the address part of the multiplexing mode information is divided into two parts and stored in the two bit positions of the address part 46B. Bits C / R and E / included in the first byte of data
A and bit FEC included in the second byte data
N, BECN, and DE are set to 0.

Thereafter, in the queue reading process 440, partial data is sequentially read from a plurality of transmission queues specified by the multiplexing mode information, and the transmission frame creation area 1 is read.
The length of the partial data to be read from each transmission queue is also specified by the multiplexing mode information. When the conditions for terminating the frame generation are satisfied, the queue reading process 440 ends.The data section 46E including a plurality of partial data (FIG. 3)
B) is generated. The details of the queue read process 440 will be described later.

After that, by the frame end processing 450,
The data part 46E (FIG. 3B) includes a frame check sequence 46F and a flag sequence 46 indicating the end of the frame.
G is added to complete one multiplex frame and transmitted.
Details of the frame end processing 450 will be described later.

Thereafter, the value of the queue status register 51 is checked (460). If there is a PPP frame in any of the transmission queues, that is, if the value of the queue status register 51 is not 0, the process 400 starts frame transmission again. The processing is started, and a new frame is generated and transmitted.
If the value of the queue status register 51 is 0, the frame transmission processing ends (470).

The queue reading process 440 is performed as follows. Referring to FIG.
A is a queue counter 5 provided in the internal register group.
The value of 4 (FIG. 7) is set to 1 (500). The first queue number and the number of read bytes are read from the multiplexing mode information selected earlier (510). It is checked whether the number of bytes of data in the transmission queue of the queue number is equal to or larger than the data of the number of bytes to be read (52).
0). If the number of data bytes in the transmission queue is insufficient, all data is read from the transmission queue (52
1), 1-byte invalid data representing the same bit string 0x-7E as the flag sequence is added by the number of insufficient bits (522). The value of the invalid byte number register 58 (FIG. 7) is increased by the number of invalid data bytes (52).
3). If there is more data than the number of bytes to be read in the transmission queue, data of the number of bytes to be read is read from the transmission queue (526).

When reading data from the transmission queue, the processor 27A stores the read data in the transmission frame creation area 1 (530), and executes the queue state update processing 1
Execute 90A. Since the number of bytes of data held in the transmission queue decreases with the reading of the data, the queue state may change. For example, when the queue status up to that time is 01, there is a possibility that the queue status of the transmission queue is changed to 00 after the above reading. Similarly, when the queue state was 10 before, the queue state may change to 01 or 00. If the previous queue state was 11, the queue state may change to 10, 01, or 00. Therefore, in the queue state update processing 190A, of the queue state update processing 190 shown in FIG. 12, the processing 310, 311 and 31 related to the change of the queue state.
Only 2, 320, 330 and 350 are executed. The multiplexing mode does not change even if the state of the transmission queue changes due to the above reading, and therefore, the length of data read from the transmission queue does not change. This is because changing the multiplexing mode increases the overhead of newly generating a frame. Therefore, the processes 340, 360, 361, and 370 are not executed. However, it is also possible to execute these processes and change the queue status update process 190A so as to change the multiplexing mode according to the change in the queue status.

The generated information length stored in the generated information length register 57 (FIG. 7) is increased by the number of read bytes (550).

Next, a frame end condition check is performed (5).
60). In this process 560, it is determined whether or not the end condition of the frame creation is satisfied. That is, referring to FIG. 17, in the frame end condition check processing 560, the processor 27A compares the generated information length in the register 57 with the maximum information length in the register 55 (561). If the transmission processing end flag is set to T
set to true (566). If not, the number of invalid bytes in the register 58 is compared with the maximum number of invalid bytes in the register 56 (562). If the number of invalid bytes is large, the transmission processing end flag is set to True (56).
6) Yes. Otherwise, the queue status register 5
It is checked whether the value of 1 is 0, that is, whether all the transmission queues are empty (563). If the value of the queue status register 51 is 0, the transmission processing end flag is set to True (566). If not, the transmission processing end interrupt flag is checked (56
4) If True, the transmission processing end flag is set to True.
e is set (566). Otherwise, the transmission processing end flag is set to False (565). When the processes 565 and 566 are executed, the frame end condition check process 560 ends.

For example, in the case of the third multiplexing mode shown in FIG.
x-80-21 is stored in the transmission frame creation area 1 (37A). Next, four bytes are read from the transmission queue 1, the transmission queue 2, and the transmission queue 4 at a time, and are written in the area following the transmission frame creation area 1 (37A). Reading is repeated until the maximum information length reaches 40 bytes. When the length of the read data reaches the maximum information length of 40 bytes, the reading process 440 ends. However, if any of the transmission queues to be read becomes empty during reading or if the data in the transmission queue becomes smaller than the number of read bytes 4, the invalid data 0x− Write 7E. Invalid data written 0x-
When the total number of bytes of 7E exceeds the maximum number of invalid bytes 4, the reading process 440 is terminated. Also, when the transmission processing end interrupt flag becomes True, the read processing 440 ends.

Here, the meaning of the above-mentioned transmission processing end condition will be described. The frame end due to the maximum information length check (561) is based on the following reason. The state of the transmission queue in use changes by reading out data for frame creation. It is desirable to use a multiplexing mode suitable for the current state of each transmission queue as much as possible. Allowing too long frames to be created using the same multiplexing mode can cause the multiplexing mode in use to be less than optimal for the actual state of the transmit queue. Therefore, the maximum information length is determined.

The end of the frame by checking the maximum invalid byte length (562) is to limit the amount of invalid data per frame so as not to lower the effective throughput. Check queue status (563)
Ends the frame generation even if the frame has not reached the maximum length and the length of invalid data in the frame is smaller than the maximum invalid byte number. Otherwise, there is a waste in adding invalid data to the frame. The check of the transmission processing end interrupt flag (564) is for terminating the generation of the frame being created so that a packet with a severe delay such as a voice packet can be immediately transmitted in the case of a waiting data transmission interrupt. . In the case of a transmission interrupt exceeding the threshold, in order to prevent the transmission queue storing packets requesting a low drop rate from overflowing, in order to increase the amount of reading from the transmission queue when the threshold is exceeded, Terminates generation of the frame being created. In this way, for packets with different requirements for delay and discard rate, processing is performed that satisfies each requirement and does not lower the effective throughput.

Returning to FIG. 16, when completing the frame end condition check, the processor 27A checks the transmission processing end flag (570). If the value of the transmission processing end flag is True, the transmission processing end interrupt flag is set to False (575), and the queue read processing ends. If the value of the transmission processing end flag is False
In the case of e, the value of the queue counter 54 is increased by 1 (580), and it is checked whether or not the value of the queue counter 54 exceeds the number of allocated queues included in the selected multiplexing mode information. (590). If the value of the queue counter 54 is larger than the number of assigned queues, the processor 27A restarts the queue reading process from the process 500. If the value of the queue counter is equal to or smaller than the number of queues assigned to the register, the process is restarted from step 510.

As described above, since the frame end condition is checked, generation of the frame may be stopped before all the packets in a certain transmission queue are included in the frame. The rest of the packet is transmitted in the next frame.

The frame end processing 450 is performed as follows. Referring to FIG. 18, the processor 27A
Calculates the frame check sequence (FCS) 46F (FIG. 3B) of the data stored in the transmission frame creation area 1 (37A) using the FCS calculation circuit 29A (FIG. 5), and calculates the transmission frame creation area 1 (37A). (451).

Next, octet stuffing processing 17
Perform 0. This process is the same as the octet stuffing process shown in FIG. As a result of the processing up to the processing 451, a plurality of partial data are concatenated, so that the data in the transmission frame creation area 1 (37A) may include the same bit string as the flag sequence. Therefore, in order to change the bit string to a bit string different from the flag sequence, the octet stuffing processing 17 is performed.
0 is executed again. This process is performed using the transmission frame creation area 1 (37A) as an input area and the transmission frame creation area 2 (37B) as an output area.

Thereafter, the processor 27A adds the flag sequence 0x- at the end of the transmission frame creation area 2 (37B).
7E (46G) is added (453). Thus, a multiplex frame is completed in the transmission frame creation area 2 (37B), and the multiplex frame is transferred to the physical layer transmission processing unit B (23).
A) (454). Thus, the frame end processing 450 ends. Physical layer transmission processing unit B (23
A) transfers the transmitted frame to the access line 3a. If there is no data to be transmitted, the flag sequence 0x-7E is transferred.

(6) Subscriber Node (6A) Device Configuration As illustrated in FIG. 19, the subscriber node 1a includes a plurality of subscriber line interfaces 14a, 14b, and 14c.
And one of the subscriber line interfaces, for example, 14a is connected to the internetwork connecting device 2a (FIG. 1) by the access line 3a. The other subscriber line interfaces 14b and 14c are connected to other inter-network connecting devices (not shown). Trunk line interface 1
3a and 13b are connected to relay nodes 5a and 5b (FIG. 1) via relay lines 6a and 6b, respectively. The switch 11 is connected to the subscriber line interfaces 14a, 14
b, 14c and switches for connecting the trunk line interfaces 13a, 13b, and the control processor 12 is a program-controlled processor for controlling the switch 11. The memory that stores the programs and data used by the control processor 12 is not shown for simplicity. Other subscriber nodes have the same structure.

When a conventional subscriber node receives a frame relay frame from an inter-network connecting device, it determines the output line of the frame according to the logical line used by the frame, and forwards the received frame there. I should have done it. However, in the present embodiment, the subscriber node 1a decomposes the multiplex frame received from the internetwork
It is configured to generate and transmit a plurality of Frame Relay frames including P packets.

As illustrated in FIG. 20, the subscriber line interface 14a has a program-controlled processor 27B and a randomly accessible memory (RAM) 28.
B is provided. The program for the processor 27B is held in the memory 28B or a ROM (not shown). The subscriber line interface 14a further includes a physical layer transmission processing unit B (23B) for transmitting a frame relay frame to the access line 3a, and a physical layer reception processing unit for receiving a frame relay frame from the access line 3a. B (24B), switch 1
1, an intra-node transmission processing unit 22B for transmitting a frame relay frame to the switch 1, and an intra-node reception processing unit 25 for receiving a frame relay frame from the switch 11.
B and a frame check sequence (FCS) calculation circuit 29B.

FIG. 21 is a diagram showing the structure of the memory 28B.
The various regions used for receiving multiple frames are shown. Receive queues 1 to 4 (65A to 65D
(FIG. 4A) is also provided in the memory 28B. Here, the number of these reception queues is, for example, four, but this number can be changed as appropriate, but the number of transmission queues 3 in the network connection device 2a corresponding to the subscriber node 1a is limited.
It is set equal to the number of 5A to 35D. The multiplexing mode information 63A to 63D held in the area 63 is set to correspond to the multiplexing mode information 33A to 33D (FIG. 6) held in the memory 28A in the network connection device 2a.

That is, as shown in FIG. 21, each multiplexing mode information stores a set of the number of the reception queue to be used and the number of bytes to be written in each reception queue.
When any of the reception queues is not used, invalid values are written in the queue number and the write byte length. Each multiplexing mode information further includes an assigned queue number indicating the total number of reception queues used in the multiplexing mode. These pieces of information are the same as the set of the transmission queue number and the number of bytes read from each transmission queue in the corresponding multiplexing mode information held in the memory 28A in the network connection device 2a, and the number of queues allocated. It is defined to have a value. The memory 2 in the network connection device 2a
The maximum information length and the maximum invalid byte number address portion in the corresponding multiplexing mode information held in 8A are stored in the memory 28B.
It is not included in the multiplexing mode information.

FIG. 22 shows examples of some multiplexing mode information, which are shown in FIG. 14 and are some of the multiplexing mode information held in the memory 28A in the interconnection device 2a. Corresponding to the example. In an initial state, the processor 27B of the subscriber line interface 14a receives the multiplex mode information from the control processor 12 and sets the multiplex mode information in the memory 28B.

The memory 28B further holds a DLCI / multiplexing mode information area conversion table 60, a routing table 61, and a queue management table 62. The queue management table 62 stores the reception queues 65A to 65D.
And has the same structure as the queue management table 32 shown in FIG. Table 60, 6
1 will be described later. The memory 28B further includes a reception frame work area 1 (66A), a reception frame work area 2 (66A).
(66B), transmission frame creation area 1 (67A), and transmission frame creation area 2 (67B).

Similarly, FIG. 23 shows data stored in the internal register group of the processor 27B.

(6B) Multiplexed Frame Reception Operation The physical layer reception processing unit B (24B)
When a frame is received from a, a reception interrupt is issued to the processor 27B. The processor 27B executes the receiving process shown in FIG. 24 in response to the interrupt. Hereinafter, only the case where the received frame is a multiplex frame will be described.

The processor 27B reads out the multiplexed frame 46 (FIG. 4A) received from the physical layer reception processing section B, and receives the received frame work area 1 (66) in the memory 28B.
A (FIG. 21)) (600). At this time, the leading flag sequence 46A of the multiplex frame 46 is deleted, but the last flag sequence 48G is left for identifying the boundary of the multiplex frame when the next frame is received during the octet destuffing process. Next, octet destuffing processing is performed (610). This process is a process of restoring the bit string changed in the octet stuffing process 170 shown in FIG.
However, for a previously received multiplex frame, the F
This octet destuffing process is suspended while the CS calculation process and the queue writing process are being performed. This processing is executed as follows using the received frame work area 1 (66A) as an input area and the received frame work area 2 (66B) as an output area. .

Referring to FIG. 25, processor 27B reads one byte of data in the input data area (70 bytes).
0), check the value. If the data read in the process 710 is other than the flag sequence 0x-7D, the read data is stored in the output area (770). if,
The data read in the process 710 is the flag sequence 0x
In the case of -7D, the processor 27B reads the next byte of the input area (720). It is determined whether or not the newly read data is 0x-5E (730). If so, the 2-byte data 0x-7D-5E read in steps 700 and 720 is converted to 0x-7E. (73
5) and store it in the output area (770). If the data read in step 720 is not 0x-5E, it is determined whether the data is 0x-5D (740). If so, read 2 bytes of data 0x
-7D-5D is converted to 0x-7D (745) and stored in the output area (770). If the data read in process 720 is not 0x-5D, the data is 0x-2
3 is determined (750). If so, convert the read 2 bytes to 0x-03 (75
5) and store it in the output area (770).

Each time the read data is stored in the output area in step 770, the processor 27B checks the end of the data in the input area by checking whether the data in the input data area is a flag sequence (780). Is not completed, the octet destuffing process is restarted from the process 700. When optional conversion is performed in the octet stuffing process 170 (FIG. 11) on the transmission side, the converted code is converted from 2 bytes to 1 byte by the octet destuffing process. If the data read in step 720 is not 0x-23, the received frame is discarded (760). If it is determined in step 780 that the input data has ended, and if the received frame has been discarded in step 760, the last flag sequence is discarded, and the octet destuffing process ends.

Returning to FIG. 24, the processor 27B determines the FCS of the received frame after the octet destuffing process 610 has been completed by the frame check sequence (F
CS) Calculation is performed using the calculation circuit 29B (FIG. 20). The FCS obtained by the calculation and the FCS (46
F) and (620). If these FCSs do not match (630), the received frame is discarded (680) because the received data is incorrect, and the frame reception process ends.

If the FCSs match (63
0), the received data is normal. Accordingly, the transmitted multiplex frame 46 (FIG. 3B) is received by the above processing. The processor 27B reads the DLCI in the address part 46B of the received frame, and uses this to search the DLCI / multiplexing mode information area conversion table 60 (FIG. 21) in the memory 28B (FIG. 20). DLCI / multiplexing mode information area conversion table 60
Includes, for each of a plurality of DLCIs used in any of a plurality of multiplexing modes, the DLCI and multiplexing mode information 63A and 63 for the corresponding multiplexing mode.
A pair with the start address of B, 63C, or 63D (FIG. 21) is stored. In the process 640, the head address of the multiplexing mode information stored corresponding to the DLCI in the received frame is read and stored in the internal register 82 (FIG. 23) in the processor 27B.

The processor 27B deletes the address part 46B from the received frame 46 (650),
2 (FIG. 23), reads out the number of assigned queues included in the multiplexing mode information having the multiplexing mode information head address, and stores the internal register 83 of the processor 27B (FIG. 23)
(660). The processor 27B performs a queue writing process (670). Queue write processing 670
Then, the plurality of partial data included in the received multiplex frame are respectively stored in the reception queues 65A and 65B in the memory 28B.
B, 65C, 65D. This is performed as follows.

Referring to FIG. 26, the processor 27B
Sets the value of the queue counter 84 (FIG. 23) for reception processing provided in one of the internal register groups to 1 (800). Next, the processor 27B reads the first pair of the queue number and the number of write bytes from the multiplexing mode information having the multiplexing mode information start address in the internal register 82 (FIG. 23), The data of the number of bytes specified by the write byte number is read and written to the reception queue specified by the queue number (810).

The processor 27B further checks whether or not unread data exists in the reception frame work area 2 (66B) (840). If there is data, the value of the queue counter 84 is increased by 1 (step 840). 85
0). The value of the queue counter 84 is compared with the number of queues allocated in the internal register 83 (FIG. 23) (860).
If the value of the queue counter 84 is larger than the number of queues allocated in the internal register 83, the queue writing process is continued from the process 800. That is, the subsequent data of the received frame is stored in the second reception queue specified by the multiplexing mode information by the number of bits specified by the information. This writing process is repeated for the value of the queue counter 84 by the number of queues allocated in the internal register 83. For example, when the third multiplexing mode information among the five multiplexing mode information shown in FIG. 22 is used, the first, second, and fourth reception queues 65A, 65A
The data of the received frame is sequentially stored in 5B and 65D in units of 4 bytes.

If the processing 80 is repeated for each repetition of writing,
If 0 is executed and the value of the queue counter 84 is determined to be larger than the number of queues allocated in the internal register 83, queue writing is repeated from step 800.
That is, a plurality of reception queues specified by the multiplexing mode information, for example, the first queue 65 of 65A, 65B, 65D.
From A, the received data is written again. Thus, the received data is sequentially divided into a plurality of partial data and stored in a plurality of reception queues. If it is determined that there is no data in the reception frame work area 2, the queue writing process is terminated (840).

In the course of the above processing, the processor 2
7B checks whether or not the written data includes the flag sequence 0x-7E every time the data received in any one of the reception queues is written by the process 810 (820). However, the flag sequence at the head of the received multiplex frame is not subject to this check. If there is a flag sequence in the written data, a packet framing process 830 is performed. If there is a flag sequence, it means that any PPP frame has been received up to the end. Therefore, in the packet framing process 830, the queue writing process is interrupted, the PPP frame having the end is restored from the data in the reception queue, and the frame in the frame relay format including the IP packet included in the PPP frame is restored. Generate and send.

Referring to FIG. 27, in the packet framing process 830, the processor 27B
The flag sequence 0x-
The 1-byte data is read from the head of the processing reception queue detected as having 7E (900). It is checked whether the read data is a flag sequence (901),
When data other than the flag sequence is read, the reception queue is set as an input area, and the transmission frame creation area 1 is set.
The octet destuffing process 610 is performed using (67A (FIG. 21)) as an output area. This processing 610 is the same as the processing shown in FIG. This process is performed to restore the code string changed by the octet stuffing process performed during the PPP encapsulation. Thus, in the transmission frame creation area 1 (67A),
The PPP-encapsulated frame 48 (FIG. 4B) is restored. The above processing is repeated until it is determined in step 901 that the flag sequence has been read.

Process 9 when the flag sequence is read
If determined at 01, the processor 27B calculates the FCS of the data in the transmission frame creation area 1 using the FCS calculation circuit 29B, and compares it with the FCS stored in the last two bytes of the transmission frame creation area 1. (920). If it is determined in process 930 that these FCSs do not match, the data stored in the transmission frame creation area 1 is discarded (970), and the packet framing process 8
End 30. When the FCSs match, the IP packet included in the PPP frame is restored as described below, and a frame relay frame including the IP packet is generated.

First, the processor 27B determines that the first three bytes of data in the transmission frame creation area 1 (67A) are P
It is confirmed that the address part 48B, the control part 48C, and the protocol part 48D (FIG. 4B) for the PP capsule (that is, 0x-FF-03-21), the bit string is discarded, and the end FCS ( 48
F) is discarded, and the IP packet 48E is taken out (94).
0). Next, the destination IP address is read from the IP header 48H, and the routing IP 61 is used as a key to search the routing table 61 (FIG. 21) provided in the memory 28B (941).

As shown in FIG. 28, the routing table 61 includes a plurality of data sets each including a destination network address, a mask, an output interface number, and an output DLCI. The processor 27B first calculates the logical product of the read destination IP address and the mask of the routing table 61, and compares the result with the destination network address of the routing table 61. Then, the output interface number of the matched row, the output DLC
I is read (942). Here, the output interface number indicates an interface to which a line for outputting a frame is connected, and the output DLCI indicates a logical line on which the frame is multiplexed on the output line.

Next, the processor 27B encapsulates the IP packet 48E in the frame relay frame 49 (950). Address part of frame header (Q.922
Address) 49B, the output DLCI read from the routing table 61 is stored in the DLCI portion.
The C / R and E / A of the first byte of the 922 address are set to 0 for FECN, BECN and DE of the second byte, and the E / A of the second byte is set to 1 (951). The processor 27B uses the FCS calculation circuit 29B,
S is calculated and stored in the FCS portion 49F of the frame 49 (952).

In this way, a frame without the last flag sequence 49G is generated from the frame 49 of the frame relay shown in FIG. 4B. Hereinafter, this frame may also be referred to as a frame relay frame for convenience.
Since the switch 11 can indicate the frame boundary by a method other than octet stuffing, for example, by another signal line, the flag sequence 49G is added by the trunk line interface 13a or 13b when the frame is transmitted later. Finally, the output interface number and the frame relay frame 49 are sent to the intra-node transmission processing unit 22B (FIG. 20) (960), and the packet framing process ends.

The in-node transmission processing section 22B sends the output interface number and frame sent to the switch 11, and
The switch 11 transmits the frame 49 to the trunk line interface 13a or 13b or the subscriber line interface 14a or 13b corresponding to the output interface number. In this case, the generated frame 49 is
Since it is assumed that the packet should be sent to a terminal (not shown) connected to the LAN 7b or 7a, the packet is sent to a trunk line interface, for example, 13a. When receiving the frame 49 from the switch 11, the trunk line interface 13a performs the same octet stuffing process as the octet stuffing process 170 on the frame 49 so that the receiving node can identify the frame. Finally, the flag sequence 49
G is added, and a complete frame relay frame is generated and transmitted to the trunk line 6a. If there is no frame to be transmitted, the flag sequence is transmitted to the trunk line 6a.

After performing the packet framing process 830, the processor 27B restarts the process 840 (FIG. 26) in the queue write process 670. When the queue writing process 670 ends, the frame receiving process ends.

(7) Transfer of Frame from Relay Network to LAN (7A) Generation of Multiple Frames at Subscriber Node In a conventional subscriber node, a frame relay frame received from another relay node is connected to it. When a logical line leading to the network connection device is used, it is sufficient to simply transfer the frame to the network connection device. However, when the subscriber node 1a receives, from the relay network 4a, a frame relay frame for a terminal (not shown) connected to the LAN 7a, the frame including the voice packet from the subscriber node 1a via the access line 3a. , A transmission delay time may occur. In order to reduce the transmission delay time, the subscriber node 1a receives a plurality of frames addressed to the terminal connected to the LAN 7a received from the relay network 4a in the same manner as the multiplexing of the frames performed by the network connection device 2a. Multiplex. This is the same for other subscriber nodes, for example, 1b and 1c.

The basic operation when the subscriber node 1a receives a frame from the relay network 4a is as follows.
In the following, the frame received by the subscriber node 1a is shown in FIG.
Assume frame 49 shown in FIG. When the subscriber node 1a receives the frame relay frame 49 from the relay network 4a, the trunk line interface 13a deletes the trailing flag sequence 49G, and performs the same octet destuffing process as the octet destuffing process 610 on the obtained frame. Do. Thereafter, the FCS (49F) in the frame 49 is checked, and if this is incorrect, the frame is discarded. FCS (4
9F) is correct, the DLCI of the address part 49B is
And the IP header 48 of the IP packet in the data section 48E
Based on the combination with the destination IP address in H, the routing table 61 (FIG. 28) is searched, the corresponding output interface number and output DLCI are read, and the output DLCI read into the DLCI area of the address portion 49B of the frame 49 is read. The frame written and obtained is sent to the switch 11 together with the read output interface number. The switch 11 transmits the frame to the trunk line interface 13a or 13b or the subscriber line interface 14a or 14b corresponding to the output interface number.

If the trunk line interface is, for example, 13a
When the frame is transmitted from the switch 11, the trunk line interface 13a performs the octet stuffing process on the frame as described above, adds a flag sequence to the end of the obtained frame, and
Send to a. If there is no frame to be transmitted, the flag sequence is transmitted to the trunk line 6a.

If the subscriber line interface, for example, 1
When a frame is transmitted from the switch 11 to 4a, the intra-node reception processing unit 25B (FIG. 20) receives the frame. Upon receiving the frame, the in-node reception processing unit 25B interrupts the reception of the processor 27B. Processor 2
7B, upon receiving a reception interrupt from the intra-node reception processing unit 25B, performs the same processing as the frame reception processing (FIG. 10) in the processor 27A of the network connection device 2a. A plurality of transmission queues are provided in the memory 28B in the subscriber line interface 14a in the same manner as the transmission queues 35A to 35D provided in the memory 28A in the internetwork connecting device 2a, and the memory 28B is provided in the memory 28A. Figure 6
Various areas for frame transmission shown in FIG. Further, the same information as the various information for transmission (FIG. 7) stored in the internal register group of the processor 27A is stored in the internal register group within the processor 27B.

However, while the processor 27A reads the frame to be received from the physical layer reception processing unit A (21 (FIG. 5)), the processor 27B reads the frame from the intra-node reception processing unit 25B. Although the frame received by the processor 27A is an Ethernet frame, the processor 27B receives a frame relay frame. Therefore, there is a difference in processing based on the difference between these frames. For example, when the processor 27B performs the process 101 in FIG.
It will be confirmed that the LPID is 0x-CC.

In this way, the processor 27B
A frame obtained by multiplexing a plurality of frames addressed to a with a PPP frame is generated. This multiplex frame is transmitted to the physical layer transmission processing unit B (23B). The physical layer transmission processing unit B (23B) transfers this frame to the access line 3a. If there is no data to be transmitted, 0x-7E is transferred.

(7B) Restoration of IP Packet and Generation of Ethernet Frame in Interworking Device The interworking device 2a receives a plurality of multiplexed frames from the access line 3a and performs the same processing as performed by the subscriber node 1a. , The IP packets are restored from these frames, and a plurality of frames including the respective packets are generated.

When receiving the frame from the access line 3a, the physical layer reception processing section B (24A (FIG. 5)) of the network connection device 2a issues a reception interrupt to the processor 27A.

When receiving a reception interrupt from the physical layer reception processing unit B (24A), the processor 27A performs the same processing as that of the multiplex frame reception processing (FIG. 24) in the processor 27B of the subscriber line interface 14a of the subscriber node 1a. Different IPs from multiple multiplexed frames
The packet is restored, a plurality of frames including each IP packet are generated, and transmitted to the LAN 7a.

For this purpose, the memory 28A in the network connection device 2a is provided with a plurality of transmission queues like the reception queues 65A to 65D provided in the memory 28B in the subscriber line interface 14a. In
Various areas for frame transmission shown in FIG. 22 provided in the memory 28B are provided. Further, a group of internal registers in the processor 27A includes a processor 27B.
Various information stored in the internal register group of FIG.
The same information as in 3) is stored.

However, since the output destination of the frame output by the processor 27A is from the relay network 4a to the LAN 7a, the frame in which the IP packet is encapsulated is not a frame relay frame but an Ethernet frame. This frame is transmitted by the physical layer transmission processing unit A (2
6A). The routing table used is changed from the table 61 in FIG. 28 to the table shown in FIG. This table is composed of a destination network address, a mask, and a destination MAC address,
From the destination IP address of the IP packet to the corresponding destination MA
The C address is searched, and the destination MAC address is set in the Ethernet frame.

As can be understood from the above description, in the present embodiment, a plurality of I
When transferring P packets, they can be multiplexed and sent, so that a packet whose transmission delay time should be reduced, such as a subsequent voice packet, can be transmitted without waiting until transmission of a large data packet is completed. Moreover, a plurality of IP packets can be classified into transmission queues based on various criteria. Therefore, an IP packet to be transferred with priority can be selected based on various criteria. Further, when multiplexing into frames, the amount (weight) of data included in the multiplexed frame can be changed for each type of IP packet. Therefore,
Multiplexing in various aspects can be realized.

Further, as information for identifying the multiplexing mode information, the DLCI of the logical line for transmitting the multiplex frame is used, and the multiplex frame is transmitted via the logical line of the DLCI. Since the DLCI is included as information originally required for transmitting the frame, the multiplexing mode used for the frame transmitted to the receiving device can be notified to the receiving device very easily. When the multiplexing mode is dynamically changed, the changed multiplexing mode can be immediately notified to the device on the receiving side. Also, it is not necessary to newly add information for notifying the multiplexing mode to the frame, and it is not necessary to notify the receiving side device by a frame different from the frame to be transmitted. Note that DLCI
Is 10 bits, so that DLCI can be assigned to a maximum of 1024 types of multiplexing modes. Furthermore, the address part of the frame relay is extended to 3 bytes and 4 bytes,
With longer DLCIs, more multiplexing modes can be used.

Using FIGS. 36A, 36B and 36C, the first
The effects of the embodiment will be described. In FIG. 36A, the subscriber line interworking apparatus 2a and the subscriber node 1a have three types of queues 1 to 3, and the following seven multiplexing modes are set as multiplexing modes of the interworking apparatus 2a. Think about it. In the multiplexing mode 1, data is repeatedly read from the queue 1 by N bytes (N is a predetermined integer) until the state of the queue changes. In the multiplexing mode 2, data is repeatedly read from the queue 2 every N bytes until the state of the queue changes. In the multiplexing mode 3, data is repeatedly read from the queue 3 every N bytes until the state of the queue changes. In the multiplexing mode 4, data is repeatedly read from the queue 1 and the queue 2 alternately by N bytes until the state of the queue changes. In the multiplexing mode 5, data is repeatedly read from the queues 2 and 3 alternately by N bytes until the state of the queue changes. In the multiplexing mode 6, data is read repeatedly from the queue 3 and the queue 1 alternately by N bytes until the state of the queue changes. Multiplexing mode 7
Reads data alternately from the queue 1, the queue 2 and the queue 3 every N bytes until the state of the queue changes. When multiplexing data in a frame in multiplexing modes 1 to 7, it is assumed that headers H1 to H7 are set in the frame, respectively. Also, the queue management table 32
By setting the wait data transmission interrupt flag, the multiplexing mode is changed when transmission wait data occurs in each queue.

Under the condition that the multiplexing mode is set as described above, as shown in FIG. 36B, a case where a packet arrives in the order of packet A in queue 1, packet B in queue 2 and packet C in queue 3 in this order. Think. In section T1, only queue 1 has data waiting to be transmitted.
As shown in section T1 of C, only the data of packet A is stored in the frame, and H1 is stored in the header of the frame. At this time, as shown in the line utilization rate of the section T1 in FIG. 36B, data of only the packet A is transmitted on the line. In section T2, the queues with data to be transmitted are queue 1 and queue 2, so as shown in section T2 of FIG. 36C, data of packet A and packet B are alternately stored in frames of N bytes each, and H4 is included in the header. Is stored. At this time, the same amount of data of packet A and packet B are transmitted on the line, and the effective band (bandwidth excluding overhead) of the line is divided into two by packet A and packet B (the line utilization rate in FIG. 36B). reference). Also, section T3
Since data waiting to be transmitted exists in the queue 1, queue 2, and queue 3, the data of the packet A, the packet B, and the packet C are alternately stored in the frame by N bytes as shown in a section T3 of FIG. Stores H7. At this time, packet A, packet B,
The same amount of data of packet C is transmitted, and the effective bandwidth of the line is divided into three by packet A, packet B, and packet C (see line utilization rate in FIG. 36B). Similarly, the section T4
In this example, the packet B and the packet C divide the effective bandwidth of the line into two, and in the section T5, the packet C occupies the effective bandwidth of the line (see the line utilization rate in FIG. 36B). (FIG. 36
In C, the header of the frame and F
Although the line overhead due to CS is shown to be constant, the line overhead actually changes because the frame length changes. In the sections T0 and T6, since there is no queue having data waiting to be transmitted, invalid data 0
Send x-7E. As described above, in the first embodiment of the present invention, the multiplexing mode of the frame is changed in accordance with the presence or absence of the data waiting to be transmitted in a plurality of queues, so that the line can be changed according to the presence or absence of the data waiting to be transmitted in each queue. Can be dynamically changed. In the example of FIG. 36, the number of bytes read from each queue is fixed to N bytes, so that the effective bandwidth is allocated evenly between queues in which data waiting to be transmitted exists. By setting to different values, it is also possible to weight the allocation of the effective band.

<Second Embodiment> In the first embodiment, multiplexing mode information that determines a multiplexing operation such as a queue to be used and a length of data to be read from each queue is stored in a memory and multiplexed. The DLCI corresponding to the conversion mode is stored in advance in the information, the length of data to be read can be specified for each transmission queue, and the value of DLCI can be determined independently of the read data length.
In the present embodiment, the queue to be used and the length of data to be read from each are not specified for each transmission queue by such multiplexing mode information, but are determined from the DLCI value used for the multiplexed frame. It has become. Therefore, this embodiment can operate with less information than the first embodiment, and requires less memory capacity.

(1) Schematic Configuration of Apparatus The DLCI included in the address part of the frame relay is composed of 10 bits as shown in FIG. In the present embodiment, as shown in FIG. 30A, a DLCI used for a multiplexed frame is divided into a plurality of regions each having 2 bits, and a value of each region is defined to specify a multiplexing method. Use For example, FIG.
As shown in the figure, the most significant 10th and 9th bits are DLC
Indicates the type of I, and when those two bits are 10, it indicates the DLCI of the frame using the multiplexing method of the present invention. When those two bits are 00, 01, and 11, they indicate a DLCI for a normal frame. In frame relay, DLCI
Are used in the user data link layer, the above two bits are used to identify the multiplex frame.

The lower 8 bits indicate a method of multiplexing a plurality of transmission queues. In this example, when two different bits make up one multiplexed frame, the length of data read from one transmission queue is shown. Here, this is also called a multiplex number, and data representing the multiplex number is also called a multiplex number code. Specifically, the eighth and seventh bits indicate the number of multiplexed transmission queues 1. Similarly, the sixth and fifth bits indicate the multiplex number of the transmission queue 2,
The fourth and third bits indicate the multiplex number of the transmission queue 3, and the second and first bits indicate the multiplex number of the transmission queue 4.

As shown in FIG. 30C, for example, when the multiplex number code corresponding to one of the transmission queues is 01, it indicates that only the N1 byte is read from the corresponding transmission queue once. The numerical value N1 is appropriately determined in advance. In the case of 10, it indicates that reading is performed N2 times N1 bytes at a time from the corresponding transmission queue. That is, as in the first embodiment, when generating a multiplexed frame, data is repeatedly read from a plurality of transmission queues, but the read length per read is N1, and this read is repeated N2 times, and Read N1 × N2 bytes. The numerical value N2 is also appropriately determined in advance. In the case of 11, it indicates that the data is read from the corresponding transmission queue by N1 bytes until the end of the frame.
00 indicates that no data is read from the transmission queue.

FIG. 31 shows a specific example of the address portion used in the multiplex frame, the value of the DLCI included in the address portion, and the corresponding multiplexing method. In the item of multiplex method,
"Until N1 ends" indicates that N1 bytes are read from the corresponding transmission queue until the end of the frame, and "no multiplex information" indicates that data is not read from the transmission queue. For example, N1 is 4 and the maximum information length is 40
When set to bytes, the multiplexing mode shown in FIG.
The multiplexing mode is the same as the multiplexing mode illustrated in FIG. 14 for the first embodiment.

Also in this embodiment, the waiting data transmission interrupt flag and the transmission interruption flag exceeding the threshold value used in the first embodiment can be set for each transmission queue. The multiplexing mode can be changed according to the amount of data to be held. For example, when data is newly set in the transmission queue in which the waiting data transmission interrupt flag is set, the code (FIG. 6C) indicating the multiplex number corresponding to the transmission queue is set to 1
Let it be 1. If the amount of data stored in the transmission queue with the threshold exceeded transmission interrupt flag set exceeds the threshold specified for that transmission queue,
Code indicating the multiplex number for the transmission queue (FIG. 6C)
Is set to 10. In other cases, the code indicating the multiplex number for the transmission queue is 01 as long as the data is held in the transmission queue. Here, it is assumed that the waiting data transmission interrupt flag and the transmission interruption flag exceeding the threshold value are not simultaneously set to the same transmission queue.

In the initial state, the rules shown in FIGS. 30B and 30C correspond to the processor 27A (FIG. 5) of the interconnection device 2a and the subscriber line interface 1 of the subscriber node 1a.
4a (FIG. 20). In the present embodiment, the memory 28B of the network connection device 2a
The area for transmission processing provided in (FIG. 5) is the same as the area shown in FIG. 6 used in the first embodiment,
In the plurality of multiplexing mode information 33A to 33D, only the address part used for the frame of the frame relay generated in each multiplexing mode is stored. The data held in the internal register group in the processor 27A of the network connection device 2a is different from that of the first embodiment, but the details are omitted from the following description because they are clear from the following description. However, in the following description, when using the same register as the register used in the first embodiment, the register used in the first embodiment is referred to.

The area provided in the memory 28B of the subscriber line interface 14a (FIG. 20) of the subscriber node 1a stores the multiplexing mode information 6 used in the first embodiment.
First only in that there is no area for 3A to 63D (FIG. 21)
This embodiment is different from the embodiment. In the present embodiment, the receiving side of the multiplexed frame does not need the multiplexing mode information, and the receiving queue to be directly stored can be determined from the received DLCI value. The data held on the internal register group in the processor 27B of the subscriber line interface 14a of the subscriber node 1a is different from that of the first embodiment, but the details are omitted from the following description because they are clear from the following description. However, in the following description, when using the same register as the register used in the first embodiment, the register used in the first embodiment is referred to.

In the present embodiment, the process of multiplexing IP packets by the network connection device 2a and transmitting a frame relay frame and the process of receiving the multiplexed frame at the subscriber node 1a are the first embodiment. It differs from the form in the following points. The generation of the multiplex frame by the subscriber node 1a and the reception of the multiplex frame by the internetwork connection device 2a are similarly changed, but the description of these operations is omitted for simplification.

(2) Multiplexed Frame Transmission Process by Inter-Network Connection Device 2a In this transmission process, processes 400, 420, 430, and 43 of the processes of FIG. 13 showing the transmission process in the first embodiment.
The following processing is performed in place of 5. In the initial state,
The processor 27A includes a maximum information length register 55 (FIG. 7), an N1 register (not shown), an N2 register (not shown), and a maximum read count register (not shown) constructed in the transmission processing area of the internal register. ) Is received from the control processor 12 of the subscriber node 1a and these registers are set. Registers not shown for the sake of simplicity are registers newly provided in the present embodiment.

When the processor 27A receives the transmission restart interrupt, the processor 27A searches the multiplex mode information area search table 3 based on the value of the queue status register 51 constructed on the internal register.
0 (FIG. 6) to find the start address of the corresponding multiplexing mode information, read the address portion from the corresponding multiplexing mode information 33A or 33D using this start address, and (37A). Next, the processor 27A extracts the DLCI from the read address portion, determines the read count threshold value for each transmission queue from the value of the DLCI, sets each of them on an internal register (not shown), The value of the read number counter (not shown) constructed on the internal register and the value of the generated information length register 57 are set to 0. The setting of the read count threshold value for each transmission queue is performed as follows.

First, 2 bits for each transmission queue are checked from the DLCI shown in FIG. 30, and if 00, the threshold value of the number of times of reading of the transmission queue is set to 0, 0
In the case of 1, it is set to 1; in the case of 10, it is set to the value N1 of an N2 register (not shown); in the case of 11, it is set to the value of a maximum read count register (not shown).

In the transmission processing according to the first embodiment, FIG.
Among the three processes, the queue read process 440 was performed according to FIG. In the present embodiment, the process 440 is executed according to FIG. The processing of FIG.
The following points are different from the above processing. Processing 51 in FIG.
0 is not executed. In the present embodiment, the length of data read from the transmission queue to be used is determined to be the number of bytes N1 indicated by an N1 register (not shown). This is because the number of times of reading is determined for each transmission queue, and the number of times is already stored in the reading number threshold register (not shown) provided for each queue as described above.

The processor 27A compares the value of the number-of-reads counter (not shown) for the transmission queue indicated by the queue counter 54 with the value of the number-of-times threshold register (not shown) for reading the queue (204).
0), if the value of the read number counter is less than the value of the read number threshold register (not shown), the data in the transmission queue of that queue number is stored in the N1 register (not shown)
It is checked whether the number of bytes is equal to or more than N1 bytes (520A). If the number of data bytes in the transmission queue is not enough, all data is read from the transmission queue (521), and 1-byte invalid data representing the same bit string 0x-7E as the flag sequence is added by the number of missing bits ( 522). Invalid byte number register 5
8 (FIG. 7) is increased by the number of bytes of invalid data to which the value of (8) is added (523). In the process 520A, when it is determined that there is data of N1 bytes or more in the transmission queue, data of the number of bytes N1 is read from the transmission queue and stored in the transmission frame creation area 1 (37A) (525).
A). After the process 523 or 525A, the read count counter (not shown) of the corresponding transmission queue is increased by 1 (2052), and the value of the generation information length register 57 is set to N.
Increment by one register (not shown) (205
5). In the process 2040, when it is determined that the value of the read count counter (not shown) of the corresponding transmission queue is equal to or more than the value of the read count threshold register (not shown), the above processes 520A, 525A, 2052, 190
A and 550A are skipped.

The queue status update process 1 is performed between the process 2052 and the process 550A and after the process 550A.
90A and a check 560A of the frame end condition are executed. The queue state update processing 190A is performed in a form in which the queue state update processing 190 (FIG. 12) when an Ethernet frame is received is changed, similarly to the queue state update processing 190A (FIG. 16) in the first embodiment. You. In the frame end condition check 560A, the frame end condition check processing 560 in the first embodiment is performed.
(FIG. 17) is executed as follows. First,
The check of the number of invalid bytes according to the process 562 shown in FIG. 17 is not performed. In the process 565, it is checked whether the value of the queue status register 51 is 0, that is, whether or not data exists in all the transmission queues. However, in the present embodiment, in this process 565, it is checked that reading from all the transmission queues has been substantially completed. That is, for all of the transmission queues with the multiplex number code of 10 where the read number N1 is specified, the read is already executed the number of times equal to the specified multiplex number N1, and the data is read until the end of the frame. It is determined that no data remains in any of the designated transmission queues of the multiplex number code 11.

(3) Multiplexed Frame Receiving Operation by Subscriber Node 1a In the initial state, the processor 27B has an N1 register (not shown) and an N2 register (not shown) provided in the reception processing area of the internal register group. (Not shown), the value of a maximum write count register (not shown) is received from the control processor 12 of the subscriber node 1a, and is set in each register. Subsequent processing is different from the operation of receiving a multiplex frame by the subscriber node 1a in the first embodiment (FIG. 24) in the following points. Instead of the process 640 in FIG. 24, the address part is read from the received frame, and the DLC therein is read.
I is taken out and DLCI built on the internal register group
It is stored in a register (not shown). The queue writing process 670 of FIG. 24 is executed according to FIG. 26, but the queue writing process 670 is executed according to FIG.
The processing of FIG. 33 differs from the processing of FIG. 26 in the following.

The processor 27B determines the threshold value of the number of times of writing for each reception queue from the value of the DLCI specified by the DLCI register (not shown), and stores the threshold in each reception queue in the reception processing area of the internal register group. Then, the value is set in a register (not shown) set for the reception queue, and the value of a write count counter (not shown) for each reception queue is set to 0 (2210). The setting of the threshold of the number of times of writing is the same as the setting of the threshold of the number of times of reading on the transmission side, and is performed as follows. First, 2 bits for each reception queue of the DLCI are checked. If 00, the value of a write count threshold register (not shown) for the reception queue is set to 0, if 01, to 1; 1
If it is 0, it is set to the value of the N2 register (not shown), and if it is 11, it is set to the value of the maximum write count register (not shown).

In FIG. 33, steps 2230, 2240, and 2250 are executed instead of step 810 in FIG. That is, the processor 27B compares the value of the write count counter (not shown) of the receive queue indicated by the queue counter 84 with the value of the write count threshold register (not shown) for the receive queue (2230), and reads out. If the value of the number-of-times counter (not shown) is less than the value of the read number threshold (not shown), the reception frame work area 2 (6
6B), the data of the number of bytes indicated by the N1 register (not shown) is read and stored in the reception queue indicated by the queue counter 84 (2240). Then, the write counter (not shown) of the corresponding reception queue is increased by 1 (2250). Subsequent processing is the same as in FIG.

In this embodiment, since the multiplexing mode is directly specified by the DLCI, the number of multiplexing modes that can be specified is limited by the number of DLCI bits and the number of queues. Also, it cannot be applied when the number of transmission queues and reception queues is very large. However, since the multiplexing mode information stored in the memory is very simple, the number of multiplexing modes may be small, which is effective for a system in which the number of transmission queues and reception queues is small.

<Embodiment 3> In this embodiment, a plurality of transmission queues are prioritized, and data belonging to a transmission queue having a higher priority is transmitted preferentially. That is, a frame including a plurality of partial data obtained by dividing the data of the transmission queue having the highest priority among the transmission queues holding the data at the time of creating the frame is generated. During the generation, when the data arrives at the transmission queue having a higher priority, the frame creation is terminated. The frames generated in this embodiment are the first and second frames.
Unlike the generation of the multiplex frame in the embodiment, the multiplex frame is not generated, but the transmission queue to be used is dynamically switched according to the arrival of the high-priority IP packet, and the The transmission delay time is first,
It can be reduced compared to the second embodiment. In the present embodiment, the logical line to be used is changed according to the transmission queue to be used, and information for identifying the transmission queue to be used is:
The DLCI of the logical line is used.

(1) Schematic Configuration of Apparatus In this embodiment, the frame transmission processing of the processor 27A in the network connection apparatus 2a and the frame reception processing in the subscriber node 1a are different from those of the first and second embodiments. different. In this embodiment, the transmission queues are assigned different priorities in ascending order of the number.
Has the highest priority, and transmission queue 1 has the lowest priority. The area for transmission processing provided in the memory 28B (FIG. 5) of the network connection apparatus 2a is the same as the area shown in FIG. 6 used in the first embodiment. 33A to 33D are transmission queues 1 and 3, respectively.
The number of one transmission queue corresponding to each of 2, 3, and 4, including the maximum information length when generating a frame to be transmitted from the IP packet in the transmission queue, and the address part of the frame. , The address section contains the DLCI to be used. In this embodiment, no multiplexed frame is generated, but the information 33A to 33D will be referred to as multiplexing mode information for convenience. The data held in the internal register group in the processor 27A is different from that of the first embodiment, but the details are omitted from the following description because they are clear. However,
In the following description, when using the same register as the register used in the first embodiment, the register used in the first embodiment is referred to.

The area provided in the memory 28B of the subscriber line interface 14a (FIG. 20) of the subscriber node 1a stores the multiplexing mode information 6 used in the first embodiment.
Although the area is the same as the area for 3A to 63D (FIG. 21), a plurality of multiplexing mode information 63A to 63D are provided corresponding to the reception queues 1, 2, 3, and 4, respectively, and each corresponding one transmission Holds the queue number. The data held on the internal register group in the processor 27B of the subscriber line interface 14a is different from that of the first embodiment, but the details will be omitted from the following description because they are clear. However, in the following description, when using the same register as the register used in the first embodiment,
Reference is made to the register used in the first embodiment.

In the present embodiment, the process of multiplexing IP packets by the network connection device 2a and transmitting a frame relay frame and the process of receiving the multiplex frame at the subscriber node 1a are the first embodiment. It differs from the form in the following points. The generation of the multiplex frame by the subscriber node 1a and the reception of the multiplex frame by the internetwork connection device 2a are similarly changed, but the description of these operations is omitted for simplification.

(2) Multiplexed Frame Transmission Process by Inter-Network Connection Device 2a In this transmission process, the process shown in FIG. 34 is executed instead of the process shown in FIG. 13 showing the transmission process in the first embodiment. Is done. The processor 27B receives the transmission restart interrupt,
From the value of the queue status register 51, a non-empty transmission queue with the highest priority is selected (1000). Next, the processor 27A reads the address portion corresponding to the transmission queue and stores it in the transmission frame creation area 1 (37A) (1010). Processes 1000 and 1010 correspond to FIG.
Are executed in place of the processes 400, 420, 430, and 435 of FIG. After that, the queue reading process is executed as follows.

The processor 27A sends 1
The byte is read and stored in the transmission frame creation area 1 (37A) (1020). Update the queue status (190
A). This processing is performed in the same manner as in the first embodiment.
This is performed by changing the queue status update process shown in FIG. The generated information length is increased by one byte (550).

Thereafter, a frame end condition check is performed as follows. It is checked whether the read one-byte data is the flag sequence 0x-7E (1030). That is the flag sequence 0x-7E
In this case, a frame end process 450 is performed. If it is not the flag sequence, it is checked based on the contents of the queue status register 51 whether or not the packet has arrived at the queue with a higher priority (1040). When the packet arrives at the queue with the higher priority, the frame end processing 450 is executed. A process 563A for checking whether the transmission queue being read has become empty is also executed. In addition, of the frame end condition check processing 560 shown in FIG. 17, the processing 561 is also executed to check whether the length of the frame being generated exceeds a predetermined maximum length. When it is determined that the end condition of the frame generation is satisfied in any of the processes 560 and 563A,
A frame end process 450 (FIG. 34) is performed.

Note that, of the processing in FIG. 17, the processing 562 is not executed. No invalid data is put into the frame. Also, the process 564 is not executed. The wait data transmission interrupt flag and the threshold over transmission interrupt flag are not used. This is because the priority is determined for each transmission queue. However, it is possible to change the present embodiment so that the priority is changed depending on the state of the queue. For example, the transmission queue in which the threshold exceeded flag is set can be given a higher priority when the threshold is exceeded.

In the process 1040 of FIG. 34, when it is determined that the packet has not arrived at the queue with the higher priority, or in the process 561, it is determined that the condition for terminating the frame generation is not satisfied. The process of 34 is a process 10
Return to 20.

The processor 27A performs the frame end processing 45
When 0 is executed, the value of the queue status register 51 is checked (460). If it is not 0, the flow returns to the process 1000 to continue the frame transmission process. If the value of the queue status register 51 is 0, the frame transmission processing is stopped (1070).

Thus, in the present embodiment, the frame generation is terminated when a flag sequence (a packet break) is found in the data of the transmitting queue (process 1030) or when the packet is transferred to the queue with the higher priority. Has arrived (1040). Therefore, except when the packet arrives at the queue with the higher priority, the frame generation ends when the last flag of the IP packet in the transmission queue being read is read. Therefore, this IP packet is transmitted while being included in one or a plurality of frames. Particularly, the case where the voice packet with the highest priority is transmitted without being interrupted in one frame increases.

(3) Reception of Multiple Frames by Subscriber Node 1a The reception of multiple frames by the subscriber node 1a is the same as that of the first embodiment shown in FIG. different. The processor 27B performs the processing 6 in FIG.
In place of 40, the DLCI is read from the received frame,
The DLCI / multiplexing mode information area conversion table 60 (FIG. 21) is searched, and a transmission queue corresponding to the DLCI of the received frame is selected (1140). The processor 27B deletes the address part from the received frame (650).
The queue write operation is executed without executing the process 660 of FIG. This queue write operation differs from the process 670 (FIG. 26) in the first embodiment in the following points. FIG.
6, the processes 800 and 850 are not executed because the queue counter is not used in the present embodiment. 26. Instead of the process 810 of FIG. 26, 1-byte data is written to the transmission queue selected in the process 1140 (1160).

As described above, in the present embodiment, the reading of the transmission queue by the internetwork connecting device 2a, the subscriber node 1
The operation of writing to the reception queue by a can be performed by reading the same transmission queue and writing to the same reception queue while the same frame is being handled, which facilitates control.

The effect of the third embodiment will be described with reference to FIG. In FIG. 37A, it is assumed that the subscriber line network connecting device 2a and the subscriber node 1a have two types of queues 1 and 2, and the priority of the queue 1 is higher than that of the queue 2. In many cases, the transfer speed of the access line 3a is lower than the transfer speed of the LAN 7a, and the reception of the next packet may end while the packet that has arrived first is being transferred. Here, the packet D arrives at the queue 2 and the access line 3a
, The packet E arrives at the queue 1 after transmission starts. When the packet E, which is a priority packet, arrives after the transmission of the packet D, which is a non-priority packet, starts, there are two conventional methods as shown in FIG. 37B. (In FIG. 37B, H indicates a frame header, and T indicates a trailer of the frame (tail control information, FCS in this embodiment). The first method (conventional method 1 in FIG. 37B) is a non-priority packet D. Is a method of transmitting the frame of the priority packet E after the transmission of the frame is completed. In the second method (conventional method 2 in FIG. 37B), the transmission of the frame of the non-priority packet D is interrupted when the priority packet E arrives, the frame of the priority packet A is transmitted, and then the frame of the non-priority packet D is transmitted. Is transmitted again. However, in the case of the conventional method 2, the first half of the non-priority packet (the part transferred before the interruption) is discarded on the receiving side, and after transmitting the priority packet, the entire non-priority packet is transmitted again as a frame.

In the conventional method 1, when the size of the non-priority packet D is large, there is a problem that the delay of the priority packet E becomes very large. Therefore, in the conventional method 2,
Although there is no delay in the priority packet E, there is a problem that the transmission of the first half of the interrupted non-priority packet D is wasted and the line utilization is reduced. On the other hand, according to the third embodiment, when the priority packet E arrives, the transmission of the non-priority packet D frame is interrupted, the priority packet E frame is transmitted, and after the priority packet E frame transmission, the non-priority packet E frame is transmitted. Since the remaining frame of D is transmitted and the non-priority packet D divided and transmitted in the receiving side queue is reconstructed, the priority packet can be added without significantly deteriorating the line utilization rate by adding only a small overhead. E can be transferred with low delay.

<Embodiment 4> In the first embodiment, a multiplexed frame is transmitted using a different logical line according to the multiplexing mode, and the receiving-side apparatus uses the DLCI in the multiplexed frame to transmit the multiplexed frame. Multiplexing modes could be identified. As described above, in the first embodiment, since the multiplexing frame has information for identifying the multiplexing mode, it is not necessary for the transmitting side to separately notify this information to the receiving side. Increase the number of The present embodiment provides a method for reducing the number of necessary logical lines without using a logical line corresponding to the multiplexing mode.

That is, in the transmission of a multiplex frame between the network connection device 2a and the subscriber node 1a, the first
Instead of using the DLCI corresponding to each multiplexing mode used in the embodiment, a logical line irrelevant to the multiplexing mode defined on the access line 3a is used. In order to notify the subscriber node 1a of the multiplexing mode from the network connection device 2a, the network connection device 2a switches the multiplexing mode and transmits a frame generated in a new multiplexing mode. Before transmission, a frame including information for identifying the used multiplexing mode is transmitted from the network connection device 2a to the subscriber node 1a.

This embodiment is different from the first embodiment in that only the part that uses the DLCI is changed in the multiplexed frame generation method according to the first embodiment. The transmission delay time of a specific IP packet such as a packet can be reduced.

In this embodiment, it is necessary to transmit the above-mentioned another notification frame, but it is sufficient to use one specific logical line for transmitting the multiplex frame.

The fourth embodiment can be applied to the second and third embodiments or various modifications to which the present invention described later can be applied. When the fourth embodiment is applied to the third embodiment, the information to be notified may be information for notifying the used transmission queue.

When the multiplexed frames having the same multiplexing mode are successively transmitted, the information to be notified may be transmitted only before transmitting the first frame.

<Embodiment 5> Instead of transmitting a notification frame in the fourth embodiment, information indicating a multiplexing mode is added to the data portion of each multiplexed frame, for example, at the head thereof. Store. Specific information indicating that this information is included may be included in the header of the frame, or the frame may be transmitted using a specially defined logical line for transmitting a multiplex frame. The receiving device is
The received multiplex frame is processed based on the information to be notified of the data part included in the multiplex frame. The determination as to whether or not the frame is a multiplex frame may be made based on the above-described specific information contained in the header portion or whether or not the frame has been transferred from the above-described specific logical line.

The fifth embodiment can be applied to the second and third embodiments or various modifications to which the present invention described later can be applied. When the fifth embodiment is applied to the third embodiment, the information to be notified may be information for notifying the used transmission queue.

As in the fourth embodiment, when the multiplexed frames having the same multiplexing mode are successively transmitted, the information to be notified may be transmitted only before transmitting the first frames. It is.

<Modifications> It is needless to say that the present invention is not limited to the above-described embodiment or its modifications, but can be realized by the following modifications or other modifications. No. Also, the present invention can be realized by a technique described as a plurality of embodiments or modifications thereof or a combination of the following modifications.

(1) The multiplexing method in the above embodiment uses a standard frame configuration, and therefore can be applied between various devices. Further, in the above embodiment, the network connection device 2a is connected to the LAN 7a and the access line 3a.
And extracts IP packets from Ethernet and Frame Relay formats, but these two frames can be any combination of frames suitable for use by the two devices sending the frame. Can be used. For example, the invention can be used for transfers between the following sets of devices:

(A) In FIG. 1, the inter-network connecting devices 2e and 2d at both ends of the trunk line 6g connecting the trunk networks 4a and 4b.
The present invention can also be applied to the transfer of an IP packet between. At this time, the transmission-side device 2e receives a plurality of IP packets encapsulated in the frame relay frame from the relay network 4a, multiplexes these IP packets into a frame relay frame, and Can be sent to

(B) The two devices to which the present invention is applied need not be directly connected by a single line. Even if a device for exchanging frames is included, the two devices are connected at the data link end point. Applicable if two devices are located. For example, in FIG. 1, the present invention can be applied to the transfer of a frame between the network connection devices 2a and 2e. In this case, the network connection device 2a can multiplex a plurality of Ethernet frames transferred from the LAN 7a to the LAN 7c, and transmit the obtained multiplexed frame to the network connection device 2e. When generating this multiplex frame, the DLCI assigned to the logical line defined between the network connection devices 2a and 2e may be used as information for identifying the multiplex mode. When relaying this multiplexed frame, the subscriber node 1a or the relay node 5b in the relay network 4a may transfer the multiplexed frame to the inter-network connecting device 2e, which is the device on the receiving side, in accordance with its DLCI without changing its contents. The network connection device 2e decomposes the multiplex frame and returns the original IP
A plurality of frame relay frames including packets may be generated and sent to the relay network 4b.

(C) The method of transferring a frame between the network connection devices 2a and 2e described in the above (b) is the same as that of the network connection device 2a corresponding to different LANs, for example, 7a and 7b.
And 2b.

(D) The network connection device 2a for the LAN 7a and the network connection device 2b for the LAN 7b are connected by a dedicated physical line (not shown), and a plurality of frame relay logical lines are defined in the physical line. In this case, the present invention can also be applied to the transfer of a frame between the network connection devices 2a and 2b.

(2) In the embodiment described above, when the IP packet is stored in the transmission queue in the network connection device 2a, it is encapsulated in PPP. In PPP, a packet included in a PPP frame can be identified in a protocol area, so that packets and frames based on various protocols can be encapsulated.

The information to be encapsulated is not limited to IP packets, but may be any frame or packet to which a data link level or packet level identifier is attached and which can be assigned to a transmission queue using the identifier. . For example, if the Ethernet frame is P
In principle, it is also possible to encapsulate a PP or a frame relay frame by PPP, and it is also possible to encapsulate a packet other than IP such as IPX.

(3) Classification of IP packets in the transmission queue can also be performed based on the destination address of the Ethernet frame or the DLCI of the frame relay.

(4) There are various methods for applying the present invention. For example, when a voice frame is transmitted by a frame relay, a different DLCI is assigned only to the frame, and the multiplexing method described in the embodiment is used in order to reduce transmission waiting in the internetwork connecting device 2a or the subscriber node 1a. It is also possible. Further, a case where both the IP and IPX protocols are supported by the output line of the multi-protocol router may be considered.

(4) The present invention can be applied to packet switching networks other than the frame relay switching network. For example, ATM
The present invention is also applicable to a packet switching network using a switch. In other words, when data belonging to a plurality of IP packets such as AAL2 is multiplexed and transmitted in the same cell in an ATM switching network using a packet switch transmitting relatively short cells, the conventional mini-cell multiplexing is used instead. Then, the partial data of each IP packet shown in the embodiment is multiplexed. No IP header is attached to each partial data. Each multiplexed cell is generated so as to have a header defined by the ATM switch. The length of the multiplex cell is also set to a fixed length determined by the ATM exchange. If PPP encapsulation is performed on IP packets prior to multiplexing, the resulting frame requires 5 bytes more than the minicell, but each multiplex cell has a plurality of headers included in the data portion of the minicell. Does not require a department. Although the final overhead ratio depends on the packet length of the IP packet, this method may reduce the increase in the overhead of the header portion due to multiplexing as compared with the case where minicells are used.

(5) In the above embodiments, each of the inter-network connection device, the subscriber node, and the subscriber line interface is constituted by a program-controlled processor, but the processing executed by the processor of each device is performed. It goes without saying that each device can be configured so that all or a part thereof is executed by a dedicated logic circuit. Further, any one of the areas for storing various information provided in the internal memory of each device, for example, the transmission queue or the reception queue, may be provided in a dedicated memory provided separately from the memory. . It goes without saying that frames or packets may be linked and stored without using a specific memory area such as a transmission queue or a reception queue.

[0223]

As is apparent from the above description, the packet transfer method according to the present invention, in which partial data of a plurality of packets is multiplexed and transmitted in one frame, transmits a specific type of packet such as a voice packet. Suitable for reducing packet transmission delay time. In addition, since the amount of packet data cut out can be changed according to the state of a plurality of packet groups, it is suitable for reducing the discard rate of specific types of packets. In addition, since a plurality of pieces of identification information corresponding to a plurality of partial data in the multiplex frame are not included in the multiplex frame, the packet transfer method according to the present invention does not significantly reduce the line efficiency.

[Brief description of the drawings]

FIG. 1 is a diagram showing an example of a packet switching network to which the present invention is applied.

FIG. 2 is a diagram schematically showing a relationship between a multiplexing mode and a logical line in the apparatus of FIG.

3A is a diagram illustrating a part of a multiplex frame transmission operation in the device of FIG. 1, and FIG. 3B is a diagram illustrating another portion of a multiplex frame transmission operation in the device of FIG.

4A is a diagram illustrating a part of a multiplex frame receiving operation in the device of FIG. 1, and FIG. 4B is a diagram illustrating another portion of a multiplex frame receiving operation in the device of FIG.

FIG. 5 is a schematic block diagram of an internetwork connecting device used in the device of FIG. 1;

FIG. 6 is a diagram showing various information stored in a memory in the network connection device.

FIG. 7 is a diagram showing various information stored in an internal register group of a processor in the network connection device;

FIG. 8 is a diagram showing a structure of a queue correspondence table used by the network connection device.

FIG. 9 is a diagram showing a structure of a queue management table used by the network connection device.

FIG. 10 is a flowchart of an Ethernet frame receiving operation by the network connection device.

FIG. 11 is a flowchart of an octet stuffing process included in the flowchart of FIG. 10;

FIG. 12 is a flowchart of a queue state update process included in the flowchart of FIG. 10;

FIG. 13 is a flowchart of an operation of transmitting a multiplex frame by the network connection device.

FIG. 14 is a diagram showing some specific examples of multiplexing mode information used by the network connection device.

FIG. 15 is a diagram showing a structure of an address portion of a frame of a frame relay.

FIG. 16 is a flowchart of a queue reading process included in the flowchart of FIG.

FIG. 17 is a flowchart of a frame end condition check process included in the flowchart of FIG. 16;

FIG. 18 is a flowchart of a frame end process included in the flowchart of FIG.

FIG. 19 is a schematic block diagram of a subscriber node used in the apparatus of FIG. 1;

FIG. 20 is a schematic block diagram of a subscriber line interface used for a subscriber node.

FIG. 21 is a diagram showing various information stored in a memory in a subscriber line interface.

FIG. 22 illustrates some specific examples of multiplexing mode information used by a subscriber line interface.

FIG. 23 is a diagram showing various information stored in a group of internal registers of a processor in a subscriber line interface.

FIG. 24 is a flowchart of an operation of receiving a multiplex frame by a subscriber line interface.

FIG. 25 is a flowchart of an octet destuffing process included in the flowchart of FIG. 24;

FIG. 26 is a flowchart of a queue writing process included in the flowchart of FIG. 24;

FIG. 27 is a flowchart of a packet framing process included in the flowchart of FIG. 26;

FIG. 28 is a diagram showing a structure of a routing table 61 used by a subscriber line interface.

FIG. 29 is a diagram showing the structure of a routing table used by the network connection device.

FIG. 30A is a data link connection identifier (DLCI) used in a packet communication network according to another embodiment of the present invention;
B is an explanatory diagram of the DLCI type designation information in the DLCI shown in A, and C is an explanatory diagram of the multiplex number designation code in the DLCI shown in A.

FIG. 31 is a diagram showing some specific examples of multiplexing mode information used in the other embodiment.

FIG. 32 is a flowchart of a queue read process executed by the network connection apparatus in the other embodiment.

FIG. 33 is a flowchart of a queue writing process executed by the subscriber line interface in the other embodiment.

FIG. 34 is a flowchart of a multiplex frame transmission process executed by the network connection device in the packet communication network according to still another embodiment of the present invention.

FIG. 35 is a flowchart of a frame reception process executed by the subscriber line interface in the still another embodiment.

36A is a diagram illustrating an example of a configuration to which the first embodiment is applied, FIG. 36B is a diagram illustrating an example of a line utilization rate according to the first embodiment, and FIG. It is a figure showing the example of.

FIG. 37A is a diagram illustrating an example of a configuration to which the third embodiment is applied, and FIG. 37B is a diagram illustrating an example of a frame transfer order in the third embodiment;

Continued on the front page (51) Int.Cl. ⁷ Identification symbol FI Theme coat II (reference) H04M 11/00 303

Claims (13)

[Claims] When a first variable length packet is received from a first network, the first variable length packet is sequentially divided and mounted on a frame, a header is added to the frame, and the frame is attached to a second network. Transmitting; receiving a second variable length packet from the first network before completing the transmission of the first variable length packet;
A predetermined length of data is cut out from the first variable length packet, a predetermined length is cut out from the second variable length packet, and the data cut out from the first variable length packet and the second variable length Multiplexing data cut out from a long packet, mounting the data on a frame, attaching a header to the frame, and transmitting the frame to the second network. 2. The method according to claim 1, wherein before the transmission of the first variable length packet is completed,
Also, when a third variable length packet is received from the first network before the transmission of the second variable length packet is completed, data of a predetermined length is cut out from the first variable length packet, A predetermined length from the second variable-length packet, a predetermined length from the third variable-length packet, data from the first variable-length packet, and data from the second variable-length packet. And multiplexing the data extracted from the third variable length packet and mounting the data on a frame, attaching a header to the frame, and transmitting the frame to the second network. The packet transfer method according to claim 1. 3. A method for separating the multiplexed data into a header attached to a frame obtained by multiplexing data cut out from the first variable length packet and data cut out from the second variable length packet. 2. The packet transfer method according to claim 1, further comprising a step of adding information and transmitting the information to the second network. 4. The information for separating the multiplexed data is separated from the frame obtained by multiplexing the data cut out from the first variable length packet and the data cut out from the second variable length packet. 2. The packet transfer method according to claim 1, further comprising transmitting the packet to the second network in a frame. 5. A method for extracting at least a part of data constituting each of a plurality of packets to be transmitted, by extracting data belonging to a header portion of each packet and data belonging to a data portion of the packet, and extracting the data from the plurality of packets. Generating a multiplexed frame having header information for transmitting the transmission data to a predetermined receiving side device, transmitting the multiplexed frame, and cutting out each packet. Changing the amount of data according to the number of the plurality of packets within a range where the length of the transmission data does not exceed a predetermined limit length. Packet transfer method. 6. The packet transfer method according to claim 5, further comprising: adding a header in which information for separating the plurality of data is written to the multiplexed frame. 7. The packet transfer method according to claim 6, further comprising the step of transmitting the information for separation in a frame different from the multiplex frame. 8. Queuing a first type of variable length packet in a first queue, queuing a second type of variable length packet in a second queue, said second queue If the second type of variable length packet is not queued in the first queue,
Reading out the variable-length packets of the first type, dividing the packets into a plurality of frames, and transmitting the divided frames. If the second type of variable-length packets are queued in the second queue, One type of variable length packet is read, a second type of variable length packet is read from the second queue, and the first type of packet and the second type of packet are multiplexed in one frame. Transmitting the packet. 9. The method according to claim 8, further comprising the step of determining whether the received packet is the first type of variable length packet or the second type of variable length packet. Packet transfer method. 10. The packet transfer method according to claim 9, wherein said second type of packet is a voice packet. 11. A processing unit for processing a packet to be transmitted to generate a frame, and a transmitting unit for transmitting the generated frame to a physical line, wherein the processing unit includes a plurality of packets to be transmitted. At least a part of the data constituting each of the packets is separated without distinguishing the data belonging to the header part of each packet and the data belonging to the data part of the packet, and transmitting a plurality of data cut out from the plurality of packets. Included as data, generates a multiplexed frame having header information for transmitting the transmission data to a predetermined receiving device, supplies the multiplexed frame to the transmission unit, and cuts out the amount of the data cut out from each packet. A packet processing device for changing the length according to the number of the plurality of packets within a range in which the length does not exceed a predetermined limit length. . 12. The packet processing apparatus according to claim 11, wherein said processing unit adds said information for separation to a header of said multiplexed frame. 13. The packet processing apparatus according to claim 11, wherein said processing unit generates a frame different from said multiplex frame, said frame including information for said separation.