JP2006254514A - Packet switching network - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a network administrating means and a packet switching apparatus for reducing delay caused by data transfer and variation of delay by avoiding the delay time caused by setting of connecions and effectively realizing the process of connectionless data flow in a large-scaled data network. <P>SOLUTION: A network is divided into a connection-oriented core network and a connectionless access network connected thereto. A plurality of connections (called permanent virtual routes (PVRs)) are extended previously to an edge node from the other edge node. The network administrating means selects one route from a plurality of PVRs and transfers data along such PVR for the connectionless data flow entering from the access network. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a network, a packet switching network, a packet switch device, and a network management device that efficiently process a large amount of connectionless data traffic using a connection-oriented network such as ATM.

  With the rapid development of the Internet in recent years, a network and a switching system that efficiently process a large amount of connectionless data traffic using a connection-oriented network such as ATM have been proposed. Note that connectionless here means that connection setting is not performed in advance prior to communication, and connection-oriented means that connection setting is required in advance prior to communication.

  For example, Non-Patent Document 1 describes the architecture of the MPOA protocol. Here, MPOA is an abbreviation for Multi Protocol over ATM (multiprotocol over ATM). When communicating by MPOA, after obtaining an ATM address converted from a layer 3 destination address (for example, destination IP (Internet Protocol) address) by the MPOA server, an ATM connection is set using an ATM signaling procedure. The ATM connection here is an SVC (Switched Virtual Connection), that is, a connection that is established on a request basis whenever there is traffic to be transmitted. A signaling procedure for handling SVC is described in Non-Patent Document 2, for example.

  Another example of the communication protocol is RSVP (Resource Reservation Protocol) described in Non-Patent Document 3. In RSVP, first, a communication band, a buffer resource of a router, etc. are reserved in order from the receiving side along the data transmission path between the transmitting side and the receiving side. Next, data transmission is started after the resource reservation is completed.

ATM Internetworking ”: Nikkei BP Publishing Center: September 22, 1995 1st edition: 121 pages "ATM Forum UNI version 3.1": Prentice-Hall, Inc .: 1995 "RSVP: A New Resource ReSerVation Protocol": IEEE Network: September 1993

  The representative of connection-oriented communication is an ordinary telephone. Real-time software processing called call admission control is required, and resources associated therewith are required. However, once resources are secured, the communication bandwidth is guaranteed. Moreover, a bidirectional band is usually secured. Of course, even if there is no sound, the resource is not released, so the resource usage efficiency is not good.

  On the other hand, connectionless communication is mainly used in a normal LAN or the like, and resources are secured and used for each data block. This is advantageous for communications in which a large amount of data is instantaneously generated in one direction, such as file transfer. However, the communication bandwidth is not always guaranteed, so if the usage rate of the entire network becomes high, resource contention will occur, and further, transfer data that could not be secured will be retransmitted. It is easy to fall into a so-called congestion state where resource shortage occurs.

  ATM has appeared to compensate for both of these drawbacks, and has contributed significantly in terms of efficient use of resources. However, on the other hand, there is also the aspect of having both drawbacks. Complex call admission control is still necessary, and if there is a shortage of resources, it will also lead to congestion.

  In addition, if all communications are switched to ATM, the advantages of ATM are fully alive and ideal, but in reality, conventional telephones are also necessary, and LAN and WAN (Wide Area Network) are widespread and easy. Will not be renewed. In particular, in the future, the use of the latter data communication will increase and traffic will increase. Therefore, efficient coexistence of these conventional data communication networks and ATM networks is required.

  Furthermore, as the name of LAN, conventional data communication is literally local communication, but recently it has been required to spread worldwide as represented by the Internet. When connecting the world with connectionless communication, a communication failure at a certain location may cause data retransmissions one after another, causing congestion all over the world instantly. Accordingly, resource management such as appropriate bandwidth allocation is necessary in a large-scale network. In addition, in order to manage a large-scale object, it is necessary to introduce a hierarchical concept.

So far, we have mainly discussed the issues related to “quantity” and “scale”, but the “quality” problem cannot be ignored. As communication becomes familiar, demands for various services are increasing. For example, in the telephone network, there are automatic transfer, caller ID notification, caller billing, and conference call service. In order to cope with this, an intelligent network for efficiently communicating control signals has been constructed. Similar demands can be expected for data communications. On the other hand, there is a plan to use an intelligent network in the same way as a telephone network, and it is also possible to logically construct a virtual control network in the ATM network by taking advantage of its features in the ATM network. . However, the consistency with the data communication network centering on the conventional LAN is not necessarily good. Therefore, in a large-scale data communication network, it is essential to grasp the traffic state, dynamically control the communication band, exchange additional information for providing services, and the like. Of course, in the “quality” problem, it is necessary to provide a mechanism in which a part of the network failure does not spread widely and a means for preventing congestion in advance.

  Hereinafter, the problems to be solved by the present invention will be described more specifically.

  When communicating with MPOA, an SVC ATM connection is established on a request basis whenever there is traffic to be transmitted, so the data transfer delay time increases due to the influence of the ATM connection setup time. In some cases, the ATM connection setup time may be longer than the data transfer time. Furthermore, since connections are established on a request basis, when the number of users who generate data increases, control packets for setting and disconnecting connections frequently occur before and after data transfer, resulting in network congestion.

On the other hand, when communicating using RSVP, since data transmission is started after resource reservation is completed, data transfer delay and delay fluctuation increase. Also, in order to continue to secure resources such as bandwidth, it is necessary to periodically send a refresh packet for retaining resources from the receiving side. Therefore, when the number of users who generate data increases, the control packet necessary for securing resources takes up a large band, and the network management itself becomes complicated.

  Therefore, an object of the present invention is to solve the problems listed below.

  First, a packet switching network, a packet switching device, and a network that avoid connection setting delays, reduce data transfer delays and delay fluctuations, and reduce the number of control packets for connection setting or resource securing It is to provide a management device.

  A second object is to provide a packet switching network, a network management device, and a packet switch device that efficiently perform connectionless data flow processing in a large-scale data network.

  A third purpose is a packet switching network and network management device that is highly resistant to physical layer failures (transmission path disconnection, etc.) and logical path failures (VC (virtual circuit), VP (virtual path, etc. disconnection)). And providing a packet switch device.

  Further, as a fourth object, packet switching for avoiding medium-term (for example, second order) and local (for example, node-by-node) congestion in a network that appears continuously due to a large amount of data, that is, burst data concentration. To provide a network, a network management device, and a packet switch device.

In the present invention, the network is divided into a connection-oriented core network and a connectionless access network connected thereto, and a plurality of connections (permanent virtual route (PVR), in advance) from the edge node to the edge node are described in this specification. Is called). The network management means selects one route from the plurality of PVRs for the connectionless data flow that has entered from the access network, and transfers data along the PVR. The state of each PVR, for example, the usable bandwidth of each PVR, is used as a route selection criterion from among a plurality of PVRs. Thus, the first object is achieved.

  As means for grasping the usable bandwidth, the network management device uses the traffic concentration grasp at each node or the bandwidth grasp by the flow control between the edge nodes using the RM packet (resource management packet). This achieves the second objective.

  A plurality of connections are set in advance, and when congestion or failure is detected, the connection is switched from the active system to the standby system. This achieves the third object.

  By adjusting (shaping) the transfer rate of the data flow to a predetermined bandwidth for each PVR at the access network side interface of the edge node, data transfer is performed along a logical route with guaranteed bandwidth.

  Further, a plurality of data links are extended from the access network to the core network using a multilink procedure, and the amount of input traffic to the core network is divided. This achieves the fourth object.

  As a first effect, setting a connection in advance eliminates the connection setting time, reduces delays and delay fluctuations associated with data transfer, and reduces the number of control packets generated for connection setting or securing resources. Can be reduced.

  As a second effect, it is possible to efficiently process a connectionless data flow in a large-scale data network by changing a route through which traffic from the connectionless access network passes according to the state of the connection-oriented core network.

  As a third effect, a packet switching network that is highly resistant to physical layer failures (transmission path disconnection, etc.) and logical path failures (VC, VP, etc. disconnection) by setting multiple connections in advance. Provided.

  As a fourth effect, a multi-link procedure is used from the access network to the core network to establish a plurality of data links, and by dividing the amount of traffic input to the core network, the medium term in the network that appears due to burst data concentration Local congestion can be avoided in advance.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

  FIG. 1 shows an embodiment of a packet switching network according to the present invention. Subnetworks #A, #B, #C, and #D are connected to edge nodes EA, EB, EC, and ED via border routers RA, RB, RC, and RD, respectively. A network area surrounded by the edge nodes EA, EB, EC, and ED is called a core network. Subnetworks connected to the core network 100 are collectively referred to as an access network. The core network and the access network are indicated by bold ellipses in the figure.

  The core network 100 is a connection-oriented network, for example, an ATM network. A data link indicated by a thin line in FIG. 1 is extended between the edge nodes EA, EB, EC, and ED and the relay nodes N1, N2, and N3. . A network management device 200 is connected to the edge nodes EA, EB, EC, ED and the relay nodes N1, N2, N3 through control signal lines (indicated by extra fine lines).

  Routing protocols in access networks such as IP (Internet Protocol) are terminated at edge nodes, and connection oriented protocols such as ATM (Asynchronous Transfer Mode) and FR (Frame Relay) operate in the core network 100, for example. is doing.

  Assume that "101.102.103.104", "104.101.102.103", "103.104.101.102" 102.103.104.101 "are assigned to the edge nodes EA, EB, EC and ED, respectively.

  FIG. 2 is a setting example of PVR (referred to as a permanent virtual route) in the packet switching network of the present invention. The network management device 200 in FIG. 1 is omitted. In the core network configuration shown in FIG. 1, PVRs R1 to R6 are registered in advance between edge nodes by the network management apparatus 200. The correspondence relationship shown in FIG. 3 exists between the PVRs R1 to R6 and the data link shown in FIG.

  FIG. 3 illustrates an example in which the ground connection information, the route, and the allocated bandwidth of the core network are managed for each PVR, corresponding to the description in FIG. The status data of FIG. 3 is registered in the network management data storage unit of FIG.

  Between the edge nodes, data is transferred along the PVR, for example, by switching with ATM cells. Between the edge node EA and the edge node EC, a plurality (two) of PVRs passing through different data links, that is, R2 and R6 are registered in advance.

  FIG. 4 is a configuration example of the network management apparatus 200 shown in FIG. The network management device 200 includes a network management control unit and a network management data storage unit. The network management control unit includes an activation unit, a data writing unit, a data analysis unit, a data collection unit, a data input unit, and a transmission / reception unit, which are interconnected by a bus.

  The transmission / reception unit is connected to each node and exchanges various state data of the edge node and the relay node. The data writing unit and the data analysis unit are connected to the network management data storage unit, and record and analyze various state data.

  Next, FIG. 21 shows a band information registration method in each node. Each route can be used from the step of measuring the dynamically changing bandwidth at each edge node or relay node, the step of notifying the measured bandwidth from each node to the network management device, and the bandwidth information sent from each node. A step of setting a bandwidth by a network management device, a step of managing a bandwidth allocated for each ground route by the network management device, a step of distributing bandwidth information from the network management device to each node, and a bandwidth information at each node It consists of steps to register.

  Each edge node and relay node in FIG. 2 also has state data. 5 to 11 are registration examples of edge nodes EA, EB, EC, ED and PVRs (permanent virtual routes) of the relay nodes N1, N2, N3 in the network configuration of FIG. Based on the management data of the network management apparatus 200 of FIG. 3, a destination subnet, VPI / VCI, and port INF (Interface) are set for each edge node.

  FIG. 5 shows a registration example of the PVR of the edge node EA. In FIG. 5, PVR-ID is a permanent route identifier, corresponds to the one described in FIG. 2, the destination subnet represents a subnet to which data from subnet A is transmitted, VPI is a virtual path identifier, This is data for specifying a virtual path for connecting each node connected to the edge node EA, here, the relay nodes N1 and N2, and the edge nodes ED and EB. Similarly to the virtual path identifier, the VCI is a virtual channel identifier that is information for specifying the virtual channel of each node connected to the edge node EA. The port INF is the port interface number of the edge node EA, and the port 1 is the port number of the interface connected to the relay node N1. FIG. 6 shows an example of registration of the PVR of the edge node EB according to the same rules as those of FIG. 5, FIG. 7 shows an example of registration of the PVR of the edge node EC, and FIG. FIG. 8 shows an example of registration of the PVR of the edge node ED.

  9 to 11 show the same rules as those of FIGS. 5 to 8, but show examples of registration of the PVR of the relay node. FIG. 9 shows the relay node N1, FIG. 10 shows the relay node N2, and FIG. Is the contents of the relay node N3.

  FIG. 12 is a configuration example of the edge nodes EA, EB, EC, and ED. The access network side line interface 10, ATM switch unit 20, core network side line interface 30, and node control unit 40 are included. The node control unit 40 is connected to the access network side line interface 10, the ATM switch unit 20, and the core network side line interface 30.

  For the ATM switch unit, for example, the common buffer switch technique disclosed in Japanese Patent Laid-Open No. Hei 4-276743 “Switching System” is used.

  FIG. 13 is a configuration example of the access network side line interface 10. A line control unit, a route table, a frame processing unit, an IP (Internet Protocol) processing unit, an AAL5 (ATM Adaptation Layer 5) processing unit, and an ATM processing unit are configured. For data transfer from the IP processing unit to the AAL5 processing unit, for example, to RFC1483 (request for comment 1483) of IETF (Internet Engineering Task Force) which is a standardization body of various Internet protocols The described multi-protocol encapsulation technique is used.

  The ATM processing unit in FIG. 13 includes a cell flow rate monitoring circuit described in Japanese Patent Laid-Open No. 4-260245 “Subscriber Line Transmission Line Capacity Display Method in Asynchronous Transfer Mode”. Measure and adjust transmission capacity.

  In the access network side interface 10 in FIG. 13, the IP connectionless data flow is converted into an ATM cell flow. FIG. 18 shows the format of an IP packet input from the access network side, and FIG. 19 shows the format of an ATM cell output to the core network side.

  FIG. 14 is a configuration example of the core network side line interface 30. It includes a line control unit, a scheduler, a cell transmission / reception unit, a cell buffer, a selector, and a physical layer processing unit. For the cell buffer, selector, and scheduler, for example, a multi-QOS shaper described in Japanese Patent Application No. 8-280860, “Flow Control Method in Network Node and Packet Switch” is used.

  Between the bandwidth information registration flow diagram (FIG. 21) and the access network side line interface diagram (FIG. 13), the portion for “measuring dynamically changing bandwidth” in FIG. Yes, the measured bandwidth is notified from each edge node to the network management apparatus 200 in FIG. 4 via the route table 13 shown in FIG. 13, the line control unit, and the node control unit 40 in FIG. The transmitted bandwidth information is temporarily stored in the data collection unit through the transmission / reception unit of the network management apparatus 200. The available bandwidth of each route is obtained by the data analysis unit of the network management apparatus 200, and set to the network management data storage unit via the data writing unit. The allocated bandwidth for each ground route is set in this network management data storage unit. to manage. This bandwidth information is transferred from the network management apparatus 200 to each edge node via the network management data storage unit, data writing unit, transmission / reception unit in FIG. 4, the node control unit 40 in FIG. 12, and the line control unit in FIG. 13 is distributed. Band information is registered in the route table 13 of each edge node.

  The operation of the network and the edge node according to the present invention will be described below in the case where data generated from the subnet #A is transferred to the subnet #C via the border router RA, the edge node EA, and the edge node EC. .

  FIG. 15 shows an example of a route table search and setting procedure in the edge node EA. R2 and R6 are set in advance as PVRs from the edge node EA to the destination subnet #C. Through the bandwidth information registration procedure shown in FIG. 21, bandwidths of 20 Mbit / s and 10 Mbit / s are set in advance for R2 and R6 of PVR, respectively.

A connectionless data flow from subnet #A to subnet #C arrives at edge node EA via border router RA. The destination subnet is determined to be subnet #C from the destination address. R2 and R6 are registered as PVRs for Subnet #C, and R2 with the largest bandwidth is selected. As a fast search method for a route having the largest bandwidth, for example, the heap sort method described on page 239 of “Data Structure and Algorithm” (Baifukan Co., Ltd .: first published in March 1987) is used.

  VPI / VCI = 11/12 and port INF = 2 are obtained as a set of VPI / VCI and port INF corresponding to R2 of the selected PVR-ID. VPI / VCI = 11/12 is set in the ATM cell header and transmitted.

  FIG. 23 shows a state where PVR is selected and data is transferred. Comparing R2 and R6 of PVR, since the bandwidth of R2 is larger than the bandwidth of R6, the first IP packet going to subnet #C entering the edge node EA is converted to an ATM cell and then transferred to R2 of PVR. It is allocated and transmitted. The numbers 1, 2, and 3 written in the ATM cell corresponding to the first IP packet indicate the transmission order of the ATM cells.

  As described above, in this embodiment, a PVR (permanent virtual route) is registered in advance in the core network 100, a destination IP address is converted into an ATM address during data transfer, and a PVR (permanent virtual route) corresponding to the ATM address is converted. Since the IP packet is transferred by selecting (route), the delay time for setting the connection is eliminated, and the delay and delay variation associated with the data transfer can be reduced. At the same time, the number of generation of control packets for connection setting or resource reservation can be reduced.

  In the core network 100, for the data to be transferred, hop-by-hop routing by processor processing (that is, the destination IP address is decoded by the processor as in the case of distributing the IP packet by the router), and the IP packet The data transfer delay in the core network is reduced because the data is loaded on the ATM cell and switched by hardware without determining the output port destination.

  In addition, by selecting the route with the highest bandwidth from a plurality of PVRs (permanent virtual routes) registered in advance in the core network 100, transfer of connectionless data flows in a large-scale data network is efficiently processed. can do.

  Next, the operation when the state in the core network changes will be described.

  FIG. 22 shows the bandwidth information at each node in the procedure shown in FIG. 21 between the network management apparatus 200 and each node in a state where IP data is transferred with R2 of PVR selected as shown in FIG. Further, it is shown that the band is re-registered and the bandwidths of 10 Mbit / s and 15 Mbit / s are newly set for R2 and R6 of PVR, respectively.

  This means that during the transfer of IP data, the state of the core network (in this example, the bandwidth) has been re-recognized by the cooperation of the network management apparatus 200 and each node.

  A connectionless data flow from subnet #A to subnet #C arrives at edge node EA via border router RA. The destination subnet is determined to be subnet #C from the destination address. R2 and R6 are registered as PVRs in the subnet #C, but R6 is selected as the PVR having the largest bandwidth.

  As a set of VPI / VCI and port INF corresponding to R6 of the selected PVR-ID, VPI / VCI = 10/17 and port INF = 1 are obtained. VPI / VCI = 10/17 is set in the ATM cell header and transmitted.

  FIG. 24 shows a state in which data is transferred after a new PVR is selected. Comparing R2 and R6 of PVR, since the bandwidth of R6 is larger than the bandwidth of R2 this time, the IP data flow going to subnet #C entering the edge node EA is converted to ATM cell and then to RVR of PVR. It is allocated and transmitted. That is, as compared with the state of FIG. 23, the route along which the ATM cell is carried is changed. The numbers 1, 2, 3, 4, 5, 6 written on the ATM cells indicate the transmission order of the ATM cells. In the state of FIG. 23, 1, 2, and 3 are ATM cells transferred via the PVR R2, and 4, 5, and 6 are transferred via the PVR after the state change, that is, the R6. ATM cells 1, 2, and 3 correspond to the first IP packet of FIG. 23, and ATM cells 4, 5, and 6 correspond to the second IP packet of FIG.

  In the case of performing route switching that makes a transition from the state of FIG. 21 to the state of FIG. 22, for example, by using an IP switch described in Japanese Patent Application No. 8-344498, Packet switching can be prevented by switching the route after passing through the cell.

  ATM cells transferred via different PVRs are assembled into IP packets by the access network side interface 10 of the edge node EC, and transferred to the subnet #C via the border router RC.

  In this way, the core network 100 periodically grasps the PVR bandwidth and distributes IP packets to an appropriate PVR according to the obtained bandwidth, thereby transferring connectionless data flows in a large-scale data network. Furthermore, it can process efficiently.

  In addition, the situation in which the control packet for setting the connection or securing the resource is generated in the core network 100 is limited to setting the PVR and updating the PVR bandwidth information every time data is transferred between the access networks. Since the connection is not always established on a request basis, the number of control packets generated in the core network can be reduced.

  FIG. 26 shows a configuration example of the AAL5 processing unit and the ATM processing unit on the input side of the access network side line interface 10 of FIG. 13 of the edge node. Input packets are input to the logical queue deployed for PVR and buffered. An output packet from the logical queue is input to the segmentation means and output as a SAR-PDU (Segmentation and Reassembly-Protocol Data Unit). Thus, the SAR-PDU output from the AAL5 processing unit (input side) 11 corresponding to PVR is input to the cell processing unit of the ATM processing unit 12 to be converted into a cell, and then input to the data rate adjustment unit. In this data rate adjusting means, shaping is performed for each PVR in accordance with the allocated bandwidth preset in the route table 13. (Shaping here means that each packet is temporarily stored in the buffer memory and the data reading speed is adjusted so as not to exceed the instantaneous data rate value.) Furthermore, the cell output order is adjusted by the server. The cell is output later.

  By shaping each PVR according to the allocated bandwidth, connectionless data flows from each subnet can be allocated to the PVR with guaranteed bandwidth. Therefore, it is possible to avoid medium-term and local congestion in the network that appears due to burst data concentration.

  In this embodiment, the usable bandwidth is used for the selection of the PVR in the edge node, but this is an example, and other information, for example, the state of the buffer of each edge node may be used. The selected route may be determined in advance according to the time, or the route may be selected at random.

  Next, the operation when congestion or failure occurs in the core network will be described.

  FIG. 17 shows a route table setting example in the edge node EA. Initially, R2 is set as the PVR going from the edge node EA to the subnet #C.

  When congestion or failure is detected in R2 as shown in FIG. 16, the route of the PVR is changed from R2 (active system) to R6 (standby system) as shown in FIG.

  Transmission line faults are detected by the ATM OAM (Operation and Management) function. The congestion is detected, for example, when a node in which congestion has occurred notifies the network management apparatus 200, and further, the network management apparatus 200 instructs each node to switch the PVR, so that the PVR is switched to the standby system.

In this way, a packet switching network that is highly resistant to physical layer failures (transmission path disconnection, etc.) and logical path failures (VC, VP, etc. disconnection) can be provided.
(Example 2)
In the first embodiment, the network management apparatus 200 intensively sets the PVR bandwidth in the route table 13 based on the bandwidth information measured at each node as shown in FIG. In addition to this, the bandwidth setting method of PVR is flow control based on ABR (Available Bit Rate) between edge nodes (Traffic management specification version 4.0 of ATM-Forum, a private standardization organization). There is a way to go and set the bandwidth. This flow control is characterized in that RM cells (resource management cells) are periodically embedded in a data cell stream to perform rate-based closed-loop feedback control. The RM cell is used as a cell for periodically monitoring an available band between edge nodes of the core network and adjusting a cell output speed according to the band.

  FIG. 27 shows a configuration example of the AAL5 processing unit and the ATM processing unit on the input side of the access network side line interface 10 of FIG. 13 of the edge node. Input packets are input to the logical queue deployed for PVR and buffered. An output packet from the logical queue is input to the segmentation means and output as a SAR-PDU (Segmentation and Reassembly-Protocol Data Unit). Thus, the SAR-PDU output from the AAL5 processing unit (input side) 11 corresponding to PVR is input to the cell processing unit of the ATM processing unit 12 to be converted into a cell, and then input to the data rate adjustment unit. In this data rate adjusting means, shaping is performed for each PVR set in the route table 13.

Next, the dynamic shaping operation for each PVR will be described.
1) In the case of explicit rate control, the maximum allowable data rate value (ACR: Allowed Cell Rate) is routed based on the explicit bandwidth information in the backward RM cell captured by the ATM processing (output side) 14 Set in table 13. The bandwidth control table 50 is set according to the ACR value, and the cell output speed is adjusted for each PVR according to the instruction. As a shaper for each PVR, for example, a common buffer technique disclosed in Japanese Patent Application Laid-Open No. 4-276743 “switching system” is used. That is, a logical queue for each PVR is formed on the common buffer, and cell reading from the common buffer is controlled using a bandwidth control table (this corresponds to a data rate adjusting means). Here, the bandwidth control table is a table for designating at which time the packet is read from the logical queue on which common buffer.

A dynamic shaping operation for each PVR can be realized by sequentially rewriting the bandwidth control table using the explicit bandwidth information in the reverse RM cell as appropriate.
2) In the case of binary mode rate control Based on the CI (Congestion Indication) bit and NI (No Increase: rate increase prohibition) bit in the reverse direction RM cell captured by the ATM processing (output side) 14, The rate calculation circuit 15 calculates a maximum allowable data rate value (ACR: Allowed Cell Rate). The difference from the case of explicit rate control is that the ACR value is relatively increased, decreased, or unchanged using the CI bit and NI bit indications. The operation after the ACR value is determined is the same as in the case of explicit rate control.

  The data rate adjusting means includes means for generating RM cells and inserting them into data.

  FIG. 20- (a) and FIG. 20- (b) show the format of the RM cell (resource management cell) used for band setting. Japanese Patent Application No. 8-51314 “Cell Output Control Circuit and Control Method” is used as a circuit for rate calculation, and means for generating RM cells and inserting them into data.

By dynamically shaping the bandwidth for each PVR reflecting the state of the core network, connectionless data flows from each subnet can be efficiently allocated to the PVR. Therefore, it is possible to avoid medium-term and local congestion in the network that appears due to burst data concentration.
(Example 3)
Figure 25 shows the use of PPP Multilink Protocol (point-to-point multilink protocol) defined in RFC1990 (Request for Comments 1990) of IETF (Internet Engineering Task Force). In this example, a data link (indicated by a bold line in the figure) is extended from the border router RA to both the edge node EA and the edge node EB. Here, the “multi-link protocol” plays a role of displaying a plurality of data links as a single logical data link. (FIG. 25 is a diagram assuming a case where the border router RA and the edge node EA and the border router RA and the edge node EB are physically connected.) In this example, it is generated from the subnet #A. The data flow of the subnet #C can be divided by the border router RA into the edge node EA and the edge node EB.

  For example, R2 and R3 are assigned as PVRs from the edge node EA to the edge node EC and from the edge node EB to the edge node EC, respectively.

  By dividing traffic using the multilink protocol in this way, medium-term and local congestion in the core network due to burst data concentration can be avoided, and traffic load balance in the core network 100 can be improved. .

  In this configuration, since a plurality of data links are extended from the border router RA to the core network 100, even if a failure occurs in the data link from the border router RA to the edge node EA, the border router RA to the edge node Since the traffic can be bypassed to the EB, fault tolerance can be increased.

It is a connection diagram which shows the structural example of the packet switching network of this invention. It is a schematic diagram which shows the example of a setting of PVR (it calls a permanent virtual route) in the packet switching network of this invention. 5 is a table showing an example of status data registered in the network management data storage unit of FIG. It is a block diagram which shows the structural example of the network management apparatus in FIG. It is a table | surface which shows the example of registration of PVR (permanent virtual route) in edge node EA. It is a table | surface which shows the example of registration of PVR (permanent virtual route) in edge node EB. It is a table | surface which shows the example of registration of PVR (permanent virtual route) in edge node EC. It is a table | surface which shows the example of registration of PVR (permanent virtual route) in edge node ED. It is a table | surface which shows the example of registration of PVR (permanent virtual route) in the relay node N1. It is a table | surface which shows the example of registration of PVR (permanent virtual route) in the relay node N2. It is a table | surface which shows the example of registration of PVR (permanent virtual route) in the relay node N3. It is a block diagram which shows the structural example of edge node EA, EB, EC, and ED in FIG. It is a block diagram which shows the structural example of the access network side line interface in FIG. It is a block diagram which shows the structural example of the core network side line interface in FIG. It is a flowchart which shows the example of the routing table search in the edge node EA, and a setting procedure. It is a schematic diagram which shows that congestion or a failure was detected by R2 in FIG. It is a flowchart which shows the example of a routing table setting in edge node EA. It is a figure which shows various data formats. It is a figure which shows various data formats. It is a figure which shows various data formats. It is a flowchart which shows the registration method of the band information in each node. It is a flowchart which shows the example of a routing table setting in edge node EA. It is a schematic diagram which shows the state in which PVR is selected and data is transferred. It is a schematic diagram which shows the state in which PVR is selected and data is transferred. It is a connection diagram which shows the other structure of the packet switching network of this invention. It is a block diagram which shows the structural example of the AAL5 process part and ATM process part of the entrance side in FIG. It is a block diagram which shows the structural example of the AAL5 process part and ATM process part of the entrance side in FIG.

Explanation of symbols

#A, #B, #C, #D ... Subnet, RA, RB, RC, RD ... Border router, EA, EB, EC, ED ... Edge node, N1, N2, N3 ... Relay node, R1, R2, R3 , R4, R5, R6 ... permanent virtual route, 100 ... core network, 200 ... network management device, 10 ... access network side line interface, 11 ... AAL5 processing (input side), 12 ... ATM processing (input side), 13 ... Route table, 14 ... ATM processing (output side), 15 ... Rate calculation circuit, 20 ... ATM switch unit, 30 ... Core network side line interface, 40 ... Node control unit, 50 ... Band control table

Claims (1)

A network for transferring packets between a plurality of packet communication networks,
A plurality of nodes that transmit and receive packets to and from the packet communication network, and a network management device that monitors and controls the state of the plurality of nodes and the network,
The network management device includes means for setting a plurality of routes in advance between a first node connected to the packet transmission source communication network and a second node connected to the packet transmission destination communication network. Have
When the first node receives the packet from the packet source communication network, the first node selects one of the plurality of routes based on the destination of the packet and the state of the network, and sends the received packet to the selected route. Having means to transfer,
The second node has means for transferring a packet from the first node to a destination communication network of the packet based on a destination of the packet.

Images