JP2002064540A - Method of packet relay processing - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a method of packet relay processing by which a minimum warrant reception band is warranted for each subscriber node and the scalability of a network is enhanced. SOLUTION: Edge nodes 1D-1F transfer packets from subscriber nodes 1J-1R to packet quality control nodes 1A, 1B and transfer packets from each packet quality control node to the subscriber nodes by referencing their destination addresses. The packet quality control nodes 1A, 1B classify the packets from the edge nodes 1D-1F by each subscriber node and transfer them to the destination addresses, limit the packet transmission band addressed to each subscriber node to a receiver side maximum transfer band of an access link accommodating the subscriber nodes, classify the packets addressed to each subscriber node by each edge node accommodating the subscriber node and transfer them to the corresponding edge node.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a packet relay processing method, and more particularly, to an IP-VPN (Internet Protocol-Vision).
The present invention relates to a packet relay processing method suitable for providing a quality control service by constructing a service network (rtual Private Network).

[0002]

2. Description of the Related Art In recent years, when a company or the like constructs a self-owned network, a virtual closed network (IP-VPN) using IP technology is being used. This allows
Compared to using a conventional dedicated line or frame relay,
Offices in various locations can be interconnected flexibly. Conventionally,
In the packet relay processing method used in the packet transfer system of such a closed network, an edge node accommodating a subscriber node refers to a destination address of an input packet and transfers the packet to the corresponding subscriber node. I was If the corresponding subscriber node cannot be identified, the packet is transferred to the packet quality control node, and the packet quality control node refers to the destination address of the packet and transfers the packet to the edge node in which the corresponding subscriber node is accommodated. Had become.

[0003]

However, in such a conventional packet relay processing method, since quality control is not performed, packets are concentrated at an edge node or a packet quality control node, and packet discard due to transmission competition. Occurred in some cases. Therefore, there is a problem that such packet discarding occurs randomly, and the minimum guaranteed reception bandwidth of a specific subscriber node cannot be guaranteed. The present invention is intended to solve such a problem, and it is possible to guarantee the minimum guaranteed reception bandwidth of each subscriber node and to improve the scalability of the network by reducing the traffic load applied to the packet quality control node. It is an object of the present invention to provide a packet relay processing method that can perform the above processing.

[0004]

In order to achieve the above object, a packet relay processing method according to the present invention provides a method for controlling packet quality at an edge node irrespective of a destination address of a packet from a subscriber node. Forwards the packet from each packet quality control node to the corresponding subscriber node by referring to its destination address, and the packet quality control node transfers the packet from the edge node to the subscriber corresponding to the destination address. In addition to classifying each subscriber node, the packet transmission band addressed to each subscriber node is limited to the maximum transfer bandwidth on the receiving side of the access link accommodating the corresponding subscriber node, and further, the packet addressed to each subscriber node is Classified by edge nodes accommodating subscriber nodes, and forwarded to corresponding edge nodes Than it is.

[0005] At this time, the edge node classifies the packet sent from the subscriber node into a plurality of logical closed groups, and determines a packet from the subscriber node belonging to the logical closed group connected to a single edge node. Is
Refers to the destination address without forwarding to the packet quality control node and forwards the packet to the applicable subscriber node. For all packets sent from other subscriber nodes, packet quality control is performed regardless of the destination address. You may make it transfer to a node. Furthermore, a plurality of packet quality control nodes are installed, and these are connected to a plurality of edge nodes, and each logical closed area group is
When a packet is transmitted from a subscriber node at an edge node, the packet quality control node is associated with the logical closed group to which the corresponding subscriber node belongs. You may make it transfer to a node.

The packet quality control node classifies the packet sent from the edge node for each subscriber node corresponding to the destination address of the packet, and also classifies the packet for each edge node accommodating the corresponding subscriber node. Grouping, weighting the packet transmission bandwidth addressed to each subscriber node included in the group for each group by the minimum guaranteed transfer bandwidth on the receiving side of the access link accommodating the corresponding subscriber node, and the packets as a whole group The transmission band may be limited to the maximum transmission band that can be simultaneously transferred to all the subscriber nodes accommodated by the corresponding edge node, and these may be transferred to the corresponding edge node.

[0007] Bandwidth observation is performed on packets from each of the subscriber nodes on the access link, and individual packets are classified into packets that fall within the minimum guaranteed transfer band and packets that exceed the minimum guaranteed transfer band, and each is marked. In the packet quality control node, each subscriber node observes the bandwidth of the packet to be received, classifies each packet into a packet that falls within the minimum guaranteed transfer bandwidth and a packet that exceeds the minimum guaranteed transfer bandwidth. When each packet is marked and the packet must be discarded due to contention of packet transmission at the edge node and the packet quality control node, the packet exceeding the minimum guaranteed transfer bandwidth is given priority by referring to the marking of each packet. You may make it discard.

The edge node assigns a priority class identifier determined in advance based on a subscription contract for each subscriber node to a packet from the subscriber node irrespective of the transmission band, so that the edge node and the packet quality can be determined. If the control node has to wait for transmission of a plurality of packets due to competition for packet transmission, the packets may be transmitted in order from the packet of the highest priority class. further,
The link bandwidth from the edge node to the packet quality control node is set to be equal to or greater than the sum of the minimum guaranteed transmission bandwidths of the subscriber nodes accommodated in each edge node, and the link bandwidth from the packet quality control node to the edge node is set to May be set to be equal to or greater than the total sum of the minimum guaranteed reception bandwidths of the subscriber nodes accommodated in the edge nodes.

[0009]

Next, embodiments of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram of a packet transfer system to which a packet relay processing method according to an embodiment of the present invention is applied. In this packet transfer system, subscriber nodes 1J, 1K, and 1L are accommodated in an edge node 1D via a physical access link 1X via an access node 1G, and the subscriber nodes 1M, 1N,
1O is accommodated in the edge node 1E by the physical access link 1Y via the access node 1H, and the subscriber nodes 1P, 1Q, 1R are accommodated in the edge node 1F by the physical access link 1Z via the access node 1I. I have.

[0010] The edge nodes 1D, 1E, and 1F are accommodated in a core relay node 1C by core links 1U, 1V, and 1W, respectively. Packet quality control node 1A,
1B is accommodated in a core relay node 1C by core links 1S and 1T, respectively. There is no particular limitation on the lower protocol layer of the link, but here all A
It follows the TM (Asynchronous Transfer Mode) protocol. When another protocol is used as the lower protocol layer of the link, the following terms related to the ATM protocol may be replaced with the corresponding terms of the other protocol.

In such a configuration, the access nodes 1G, 1H, and 1I as calling devices operate as shown in FIG. FIG. 2 shows a configuration example of the calling-side access node. Packets from the subscriber nodes 2F, 2G, and 2H are received from the input port 2D, and are distributed to the per-subscriber queue 2B. The packets stored in the queue for each subscriber are transmitted by the scheduler 2C in ATM cell units. At this time, the bandwidth is shared between the subscriber nodes. That is,
If there is no transmission contention, packet transfer using all available bandwidths is possible, but if there is transmission contention, the consumable bandwidth is equally divided among the subscriber nodes.
However, it is also possible to perform weighting according to the contract.

As a scheduling algorithm in the scheduler 2C, a WRR (Weig
hted Round Robin). Thereby, for each subscriber node, the maximum transfer band is defined as the transfer band when there is no transmission contention, and the minimum guaranteed transfer band is specified as the minimum consumable transfer band even when there is transmission contention. . The scheduler 2C further includes a packet corresponding to the range of the minimum guaranteed transfer bandwidth,
Tagging is performed on the CLP field of the ATM cell in order to distinguish the packet corresponding to the excess of the minimum guaranteed transfer band. That is, the header part CL of the ATM cell constituting the packet corresponding to the range of the minimum guaranteed transfer band.
"0" is described in the P field, and "1" is described in the header section CLP field of the ATM cell constituting the packet corresponding to the excess of the minimum guaranteed transfer bandwidth. This CLP value is hereinafter referred to as a band identifier.

[0013] The scheduled packet is output from the output port 2E, and is transferred to the edge node 2I. Between the access node and the subscriber node,
As a logical access link, GFR (Generic Flow
Rate) class, an ATM-VC connection is set. The maximum transfer band and the minimum guaranteed transfer band are GFR
Defined as class ATM-VC bandwidth quality.

An edge node 1D as a source device, which has received a packet from an access node 1G, 1H, 1I,
1E and 1F operate as shown in FIG. FIG. 3 shows a configuration example of the originating edge node. When a packet is received from the input port 3L, the packet is transferred to the VPN separation unit 3B to perform VPN identification. The VPN identification is based on the VC described in the header of the ATM cell constituting the input packet.
By referring to I, each subscriber node is identified, and contract VPN information of each subscriber node is specified.
At this time, contract class information is also specified at the same time.

The packets for which VPN identification has been performed are VPN-specific packet forwarders 3C, 3D,
Sent to 3E. VPN dedicated packet forwarder 3C,
In 3D and 3E, a packet transfer process in the IP layer is performed, and a label is given based on the result.
FIG. 4 shows a configuration example of the core frame. The labeled IP packet is called a core frame. As shown in FIG. 4, not only the destination edge node identifier 4A is described for the label, but also the class specified earlier together with the VPN identifier 4B. The identifier 4C is also described at the same time. In this class identifier, in addition to the contract class, the band identifier described in the CLP field of the cell header is also described.

Here, the destination of a core frame generated by a packet from a subscriber node belonging to a VPN belonging to a band guarantee type class and distributed and accommodated by a plurality of edge devices is predetermined for each VPN. Packet quality control node. Further, the destination of the core frame generated by the packet from the subscriber node belonging to the band non-guaranteed type class is the edge node accommodating the destination subscriber accommodation node. Further, the destination of a core frame generated from a packet from a subscriber node belonging to a VPN belonging to a band guarantee type class and centrally accommodated in a single edge device is the self-edge node.

The IP header 4D and the payload 4E are not rewritten and are transferred transparently. The core frame generated by the labeling is sent to the core frame forwarder 3F. The core frame forwarder 3F is shared between VPNs. The core frame forwarder 3F specifies a transfer route of the core frame based on the destination edge node identifier of the label, and specifies an output port in the edge node. The core frame output from the output port is transferred to the class separation unit 3G in advance, and the class dedicated queue 3
H, 3I and 3J are separated and accumulated. The core frames stored in these queues are subjected to scheduling according to the priority of the class by the scheduler 3K, and are output from the output port 3M. Edge node 1
The core frames output from D, 1E, and 1F are delivered to core relay node 1C via core links 1U, 1V, and 1W.

When receiving the core frame from the edge node, the core relay node 1C operates as shown in FIG. FIG. 5 shows a configuration example of the core relay node. Core relay node 5
A receives the core frame from the input port 5H,
This is transferred to the core frame forwarder 5B. In the core frame forwarder 5B, the transfer direction of the core frame is specified based on the destination edge node identifier of the label, and the output port 5I in the edge node is specified. The core frame output from the output port 5I is transferred to the class separation unit 5C in advance, and the class-dedicated queues 5D, 5E, 5F
Separated and accumulated.

The core frames stored in these queues are subjected to scheduling according to the priority of the class by the scheduler 5G, and are output from the output port 5I. At this time, when a large number of core frames are accumulated in a single queue, and the accumulated number exceeds a predetermined threshold, the core frame with the bandwidth identifier “1” is discarded preferentially, and the bandwidth is discarded. The discard of the core frame whose identifier is “0” is prevented as much as possible. The core frames output from the core relay node 1C include the core links 1S, 1T, 1U, 1
V, 1W, and transferred to the edge nodes 1D, 1E, 1F or the packet quality control nodes 1A, 1B.

When receiving the core frame, the packet quality control nodes 1A and 1B operate as shown in FIG. FIG.
2 shows a configuration example of the packet quality control node. When receiving the core frame from the input port 6P, the packet quality control node 6A transfers this to the core frame forwarder 6B. The core frame forwarder 6B performs VPN separation based on the VPN identifier described in the label of the packet, and transfers the packet with the label removed to the VPN-dedicated packet forwarders 6C and 6D. At this time, the band identifier described in the class identifier is added to the packet as an attribute.

VPN dedicated packet forwarders 6C and 6D
Receives a packet, separates the packet for each destination subscriber node based on the destination IP address of the packet, and separates the subscriber-specific queues 6E, 6F, 6G, 6H, 6H.
Transfer to I, 6J. Subscriber-only queues 6E, 6F, 6
The packets transferred to G, 6H, 6I, 6J are stored here. At this time, if a large number of packets are accumulated in a single queue and the accumulated number exceeds a predetermined threshold, packets with a bandwidth identifier of “1” are preferentially discarded, and a bandwidth identifier of “1” is set. Discarding the packet of "0" is prevented as much as possible. Subscriber-only queues 6E, 6F, 6G, 6H,
6I and 6J are stored in the scheduler 6
The scheduling process is performed by K and 6L. Here, band shaping is performed using the maximum reception band of the receiving access link of the subscriber node.

The schedulers 6K and 6L further distinguish packets corresponding to the range of the minimum guaranteed reception bandwidth of the access link accommodating the destination subscriber node from packets corresponding to the excess of the minimum guaranteed reception bandwidth. Next, re-tagging is performed on the band identifier in the class identifier of the packet label. For the scheduled packet, the destination edge node is specified by the VPN-dedicated packet forwarders 6M and 6N, and the label is re-assigned based on the result. The core frame provided with the label is sent to the core frame forwarder 60.

The core frame forwarder 60 specifies a transfer route of the core frame based on the destination edge node identifier of the packet label, and specifies an output port in the edge node. If necessary, the same scheduling processing as that of the core relay node is performed. The core frames transmitted by the packet quality control nodes 1A and 1B are transmitted via the core relay node 1C to the destination edge nodes 1D, 1E,
Delivered to 1F.

Edge nodes 1D and 1 as destination devices
Upon receiving the core frame, E and 1F operate as shown in FIG. FIG. 7 shows a configuration example of the destination edge node. When receiving the core frame from the input port 7L, the edge node 7A transfers the core frame to the core frame forwarder 7B. The core frame forwarder 7B performs the VPN separation based on the VPN identifier described in the label, and removes the label-eliminated packet from the VPN packet forwarder 7C.
Transfer to 7D and 7E. At this time, the class identifier is added to the packet as an attribute.

The VPN-dedicated packet forwarders 7C, 7
In D and 7E, the ATM-VC as the logical access link identifier to output the packet is obtained from the destination IP address.
Specify I (Virtual Channel Identifier). Thereafter, the packet is divided into ATM cells and the cell forwarder 7
Transfer to F. At this time, a band identifier extracted from the class identifier of the packet attribute is described in the CLP field of the cell header. The cell forwarder 7F specifies a cell transfer route based on the CLP of the cell header, and specifies an output port in the edge node. The cells output from the output port are transferred to the class separation unit 7G in advance, separated and stored in the class dedicated queues 7H, 7I, 7J.

The cells stored in these queues are subjected to scheduling according to the priority of the class by the scheduler 7K and output from the output port 7M. The cells output from the edge nodes 1D, 1E, 1F are delivered to the access nodes 1G, 1H, 1I via the physical access links 1X, 1Y, 1Z.

Access node 1G, 1 as destination device
H and 1I receive the cell from the edge node,
Works like FIG. 8 shows a configuration example of the destination access node. Upon receiving the cell from the input port 8D, the access node 8A transfers the cell to the destination subscriber separating unit 8C. In the destination subscriber separating unit 8C, the VCI of the cell header is
The cell is distributed to the per-subscriber queue 8B with reference to the (Virtual Channel Identifier). The cells stored in the per-subscriber queue 8B are transmitted from the output port 8E to the subscriber node 8
G, 8H, 8I. At this time, when a large number of packets are accumulated in a single queue and the accumulated number exceeds a predetermined threshold value, packets with cell header CLP of “1” are discarded preferentially, and cell header CLP is set to “0”. "
As much as possible. The cell output from the access node is received by the subscriber node.

In the packet transfer system of FIG. 1 described above, the band of the core link 1U is the sum of the minimum guaranteed transfer bands of the logical access links connecting the subscriber nodes 1J, 1K, 1L accommodated in the edge node 1D. And the bandwidth of the core link 1V is set so as not to fall below the sum of the minimum guaranteed transfer bandwidths of the logical access links connecting the subscriber nodes 1M, 1N, and 10 accommodated in the edge node 1E. I do.

Further, the bandwidth of the core link 1W is determined by each of the subscriber nodes 1P, 1Q, 1R accommodated in the edge node 1F.
Is set so as not to be less than the total of the minimum guaranteed transfer bandwidth of the logical access link connecting the logical access link, and the bandwidth of the core link 1S is set to the logical access connecting each subscriber node belonging to the VPN accommodated in the packet quality control node 1A. Set so that it does not fall below the sum of the minimum guaranteed transfer bandwidth of the link. Furthermore, the band of the core link 1T is set so as not to be less than the total minimum guaranteed transfer band of the logical access link connecting each subscriber node belonging to the VPN accommodated in the packet quality control node 1B.

In the above description, the packet quality control nodes are distributed for each VPN. However, when applying the invention of claim 4, it is necessary to distribute the packet quality control nodes for each edge node. In this case, the packet quality control node is selected for each VPN at the edge node, but the packet quality control node is selected for each edge node regardless of the VPN. Also, in the packet quality control node, the bandwidth shaping is performed by the maximum reception bandwidth of the receiving access link of the subscriber node for each subscriber. WRR in cell units, similar to the originating access node
Perform scheduling equivalent to scheduling. Further, the setting of the bandwidth value may be set to the same value based on the maximum transfer band and the minimum guaranteed transfer band on the receiving side in the access link of the corresponding subscriber in the per-subscriber node queue.

As described above, according to the present invention, all packets are always transferred via the packet quality control node, and the packet quality control node performs the scheduling for each subscriber node. A guaranteed reception band can be guaranteed. Also, in all the subscriber nodes, the packets of the logical closed group connected to a single edge node are forwarded back at the edge node, so that the packet quality control node performs scheduling for each subscriber node. In this case, it is possible to reduce the traffic load applied to the packet quality control node while guaranteeing the minimum guaranteed reception bandwidth of each subscriber node by utilizing the fact that the return traffic and the traffic from the packet quality control node do not interfere with each other. Therefore, even if only a single packet quality control node is used, the entire network can accommodate more subscribers.

Further, since the packet quality control nodes are distributed in several VPN units, even if the processing capability of a single packet quality control node is limited,
The entire network can accommodate more subscribers. In addition, since the packet quality control nodes are distributed in units of several edge nodes, even if the processing capacity of a single packet quality control node is limited,
The entire network can accommodate more subscribers. Also, discard control is performed by discriminating packets that fall within the minimum guaranteed transfer bandwidth and packets that exceed the minimum guaranteed transfer bandwidth. Packets exceeding the minimum guaranteed transmission bandwidth of the subscriber router can be preferentially filled from packets that fit within the minimum guaranteed transmission bandwidth, increasing the fairness between each sending subscriber router. .

Further, by providing a plurality of priority classes at the edge node and performing priority control at the edge node and the packet quality control node, various quality classes can be prepared. The service can be provided by mixing a quality class that can guarantee the minimum guaranteed reception band and a quality class that does not guarantee the reception band at all. Further, in the relay link between the edge node and the packet quality control node, a link bandwidth set to be equal to or greater than the sum of the minimum guaranteed transmission bandwidths of the subscriber nodes accommodated in the edge node is secured, so that each The minimum guaranteed reception bandwidth of the subscriber node can be guaranteed.

[0034]

As described above, according to the present invention, the edge node transfers all the packets from the subscriber node to the packet quality control node regardless of the destination address, and transfers the packet from each packet quality control node. Refer to the destination address and forward to the applicable subscriber node,
The packet quality control node classifies packets from the edge node into each subscriber node corresponding to the destination address, and sets the packet transmission band addressed to each subscriber node to the receiving side of the access link accommodating the corresponding subscriber node. , And further classifies packets addressed to each subscriber node into edge nodes accommodating the corresponding subscriber node, and transfers them to the corresponding edge node.

Therefore, all packets are always transferred via the packet quality control node, and the packet quality control node performs scheduling for each subscriber node, so that the minimum guaranteed reception bandwidth of each subscriber node can be guaranteed. In addition, the traffic load applied to the packet quality control node can be reduced, and the scalability of the network can be improved. In addition, fairness among the sending subscriber routers can be ensured, and various quality classes can be prepared. When an ATM network or a frame relay network is used in the lower layer, a greater effect can be obtained.

[Brief description of the drawings]

FIG. 1 is a block diagram of a packet transfer system to which a packet relay processing method according to an embodiment of the present invention is applied.

FIG. 2 is a block diagram illustrating a configuration example of an originating access node.

FIG. 3 is a block diagram illustrating a configuration example of a source edge node;

FIG. 4 is an explanatory diagram showing a configuration example of a core frame.

FIG. 5 is a block diagram illustrating a configuration example of a core relay node.

FIG. 6 is a block diagram illustrating a configuration example of a packet quality control node.

FIG. 7 is a block diagram illustrating a configuration example of a destination edge node.

FIG. 8 is a block diagram illustrating a configuration example of a destination access node.

[Explanation of symbols]

1A, 1B packet quality control nodes, 1C ... core relay nodes, 1D, 1E, 1F ... edge nodes, 1G, 1H,
1I access node, 1J, 1K, 1L, 1M, 1
N, 1O, 1P, 1Q, 1R ... subscriber node, 1S, 1
T, 1U, 1V, 1W ... core link, 1X, 1Y, 1Z
... physical access link, 2A ... access node, 2B ...
Queue per subscriber, 2C scheduler, 2D input port, 2E output port, 2F, 2G, 2H subscriber node, 2I edge node, 3A edge node, 3B
VPN separation unit, 3C, 3D, 3E: VPN dedicated packet forwarder, 3F: Core frame forwarder, 3G: Class separation unit, 3H, 3I, 3J: Class dedicated queue, 3
K: scheduler, 3L: input port, 3M: output port, 4A: destination edge node identifier, 4B: VPN identifier, 4C: class identifier, 4D: IP header, 4E: payload, 5A: core relay node, 5B: core frame Forwarder, 5C class separation unit, 5D, 5E5F class dedicated queue, 5G scheduler, 5H input port, 5I output port, 6A packet quality control node, 6B core frame forwarder, 6C, 6D VP
N dedicated packet forwarder, 6E, 6F, 6G, 6H,
6I, 6J: subscriber-only queue, 6K, 6L: scheduler, 6M, 6N: VPN-only packet forwarder, 6
O: core frame forwarder, 6P: input port, 6Q
... Output port, 7A ... Edge node, 7B ... Core frame forwarder, 7C, 7D, 7E ... VPN dedicated packet forwarder, 7F ... Cell forwarder, 7G ... Class separation unit, 7H, 7I, 7J ... Class dedicated queue, 7K ... Scheduler , 7L ... input port, 7M ... output port, 8A
... access node, 8B ... per-subscriber queue, 8C ... destination subscriber separation unit, 8D ... input port, 8E ... output port,
8F: edge node, 8G, 8H, 8I: subscriber node.

 ──────────────────────────────────────────────────続 き Continuing on the front page (72) Keiichi Hoshi 2-3-1, Otemachi, Chiyoda-ku, Tokyo Nippon Telegraph and Telephone Corporation (72) Hiroyuki Hara 2-3-3, Otemachi, Chiyoda-ku, Tokyo No. 1 Nippon Telegraph and Telephone Corporation F-term (reference) 5K030 HA08 HD03 KX29 LC18

Claims (7)

[Claims] An access link in which a maximum transfer band and a minimum guaranteed transfer band are defined on both a transmitting side and a receiving side is used to connect a plurality of subscriber nodes to an edge node for accommodating the subscribers. A packet relay processing method used in a packet transfer system in which a plurality of edge nodes and a packet quality control node for relay are connected using a relay link, wherein a packet from a subscriber node is sent to a destination at an edge node. All packets are transferred to the packet quality control node irrespective of the address, and the packets from each packet quality control node are transferred to the corresponding subscriber node by referring to the destination address. Are classified for each subscriber node corresponding to the destination address, and packet transmission destined for each subscriber node is performed. Restrict the bandwidth to the maximum transfer bandwidth on the receiving side of the access link accommodating the relevant subscriber node, further classify packets addressed to each subscriber node for each edge node accommodating the relevant subscriber node, and divide them. A packet relay processing method, wherein the packet is forwarded to a corresponding edge node. 2. The packet relay processing method according to claim 1, wherein the edge node classifies the packet sent from the subscriber node into a plurality of logical closed grooves, and the logical node connected to a single edge node. Packets from subscriber nodes belonging to the closed group are forwarded to the corresponding subscriber node by referring to the destination address without being forwarded to the packet quality control node, and are sent from other subscriber nodes. A packet relay processing method, wherein all packets are transferred to the packet quality control node regardless of the destination address. 3. The packet relay processing method according to claim 1, wherein a plurality of packet quality control nodes are installed, and these are connected to a plurality of edge nodes, and each logical closed group is defined as A packet quality control node associated with the logical closed group to which the corresponding subscriber node belongs when a packet is transmitted from the subscriber node at the edge node. A packet relay processing method, wherein the packet is transferred to a node. 4. The packet relay processing method according to claim 1, wherein the packet quality control node classifies a packet sent from the edge node for each subscriber node corresponding to a destination address of the packet, Grouping is performed for each edge node accommodating the corresponding subscriber node, and for each group, the packet transmission bandwidth addressed to each subscriber node included in the group is determined on the receiving side of the access link accommodating the corresponding subscriber node. Weighted by the minimum guaranteed transfer bandwidth, the packet transmission bandwidth of the entire group is limited to the maximum transmission bandwidth that the corresponding edge node can simultaneously transfer to all the subscriber nodes accommodated by the edge node, and these are set to the corresponding edge nodes. A packet relay processing method characterized by sending back. 5. The packet transfer system according to claim 1, wherein a band observation is performed on a packet from each subscriber node on the access link, and a packet that fits in a minimum guaranteed transfer band and a minimum guarantee packet are obtained. The packet quality control node performs a band observation on a packet to be received by each of the subscriber nodes, classifies the packet as a packet exceeding the transfer band, performs a marking on each packet, and places the individual packet within the minimum guaranteed transfer band. If the packet must be discarded due to contention of packet transmission at the edge node and the packet quality control node, the marking is performed on each packet. And preferentially discard packets that exceed the minimum guaranteed transfer bandwidth A packet relay processing method. 6. The packet transfer system according to claim 1, wherein the edge node previously determines a packet from the subscriber node for each subscriber node based on a subscription contract regardless of a transmission band. When the edge node and the packet quality control node wait for transmission of a plurality of packets due to contention of packet transmission, the packet is transmitted in order from the packet of the highest priority class. Characteristic packet relay processing method. 7. The packet relay processing method according to claim 1, wherein a link bandwidth from the edge node to the packet quality control node is a minimum guaranteed transmission bandwidth of a subscriber node accommodated in each edge node. And a link bandwidth from the packet quality control node to the edge node is set to be equal to or greater than a minimum guaranteed reception bandwidth of a subscriber node accommodated in each edge node. Relay processing method.