JP6082330B2 - Communication apparatus, program, and method for increasing accommodation efficiency of aggregated packets for each path in consideration of allowable waiting time - Google Patents

Description

  The present invention relates to a technology of a communication apparatus having a different queue buffer for each path in order to transfer an input packet to a different path. In particular, the present invention relates to a technology of an edge device that communicates via a core network.

  The core network includes a core device specialized for packet transfer processing and an edge device that selects a packet route. The core device is arranged at the core of the core network, and the edge device is arranged at the boundary between the core network and the access network. In particular, the edge device grasps the local availability in the core network and selects a packet transfer path according to the communication quality required for each packet. Since the edge device has a larger processing load than the core device, a plurality of edge devices are often provided at one base and are distributed in load.

For example, a communication device such as an edge device needs to establish the following relationship in order to make maximum use of the entire accommodated line of the core network.
Allowable average packet length ≥ Ml / Mp
Ml: number of bits of the accommodated line per second in the communication device Mp: number of packets that can be processed per second in the communication device

  FIG. 1 is a system configuration diagram for communicating between edge devices via a core network.

  According to FIG. 1, the transmitting edge device establishes a plurality of different paths for the same destination edge device via the core network. Then, the transmission-side edge device sends out the packet input from the access network to a route having a free communication band as much as possible. As a result, the surplus communication path is effectively used to increase the utilization efficiency of the entire core network.

  In addition, since the edge device includes a separate queue buffer for each path, a communication delay of the queue buffer with respect to an input packet is also important. The edge device needs to optimally configure the queue buffer in order to perform relay transfer according to the priority (communication quality) of each received packet. Specifically, the packet is relayed and transferred in consideration of the waiting time in the buffer measured for each packet and the allowable waiting time based on the priority. According to the core network that transfers packets of different services, different communication delays are required for each service. This communication delay is given as the time that the packet can stay in the edge device, that is, the allowable waiting time that can stay in the queue buffer. The edge device transfers the packets to the core network while satisfying the communication delay constraint.

  Further, the edge device has a flow table that stores the address group of the destination edge device and its route in association with each other. According to the flow table, a route for each packet, that is, a queue buffer to which the packet is to be input is selected. Specifically, different queue buffers are selected also by a routing control protocol such as RIP (Routing Information Protocol), OSPF (Open Shortest Path First), and BGP (Border Gateway Protocol). The path between the edge devices may be different physical lines, or may be a plurality of logical lines set in a single physical line.

  As another conventional technique, there is a technique of aggregating a plurality of packets input from a radio channel for a radio base station and outputting the aggregated packet to a fixed network (see, for example, Patent Documents 1 and 2). According to this technique, the CPU waits until a plurality of packets are stored in the queue buffer, aggregates the packets in the queue buffer at a predetermined timing, and outputs the aggregated packets. Even when the average packet length of packets to be transferred is smaller than the allowable average packet length, the average packet length of those packets can be increased. By transferring in units of packets having a large average packet length, the network utilization efficiency can be increased.

JP 2012-239243 A JP 2013-102433 A

According to the prior art described above, it may be difficult to pursue both of the following.
(1) Maximize network utilization based on route selection for each packet (2) Maximize allowable average packet length based on aggregation of multiple packets

(1) Maximizing network utilization based on route selection for each packet Normally, when a packet is distributed to a plurality of routes, a packet distribution rate (distribution rate) is given for each route. The edge device distributes each packet to a different route according to the distribution rate. Packets are input to the queue buffer of the route with a high distribution rate at a higher rate than the queue buffer of the route with a low distribution rate.

  Note that this distribution rate is periodically updated in order to cope with the usage rate of the high-speed network that changes from moment to moment and the change in the traffic volume input to the edge devices surrounding the high-speed network. According to FIG. 1, the management device 2 is provided in the core network, and the distribution rate for each queue buffer is periodically transmitted to the edge device.

(2) Maximum allowable average packet length based on aggregation of multiple packets When packets that require a certain communication quality (communication delay) are aggregated, packets with a short allowable waiting time can stay in the queue buffer The time is short. When such a packet is distributed to a route with a low distribution rate, the queue buffer has a shorter waiting time for a subsequent packet for aggregation, and the queue buffer does not accumulate enough packets. The stored packet is output. According to FIG. 1, the management apparatus 2 transmits to the edge apparatus 1 an allowable standby time corresponding to a destination IP address and a flow tuple (a pair of a source / destination IP address and a port number). It may be.

  Therefore, an object of the present invention is to provide a communication device, a program, and a method that can increase the accommodation efficiency of aggregated packets for each path in consideration of an allowable waiting time.

According to the present invention, in order to transfer an input packet to a different route, in a communication device having a different queue buffer for each route,
The input packet is given an allowable waiting time dp,
For each queue buffer i, the path bandwidth w _i in the path output from the queue buffer i is set,
The queue buffer accommodates and outputs a plurality of packets in one aggregated packet when the allowable waiting time dp is reached for the accumulated packets,
Among the plurality of queue buffers, a queue buffer having a narrower route bandwidth w _i has a route selection means for inputting a packet having a larger allowable waiting time dp.

According to another embodiment of the communication device of the present invention,
For each queue buffer i, an average allowable waiting time dave _i for a plurality of packets input in the past is further recorded,
The route selection means
For each queue buffer i, a deviation g _i between the average allowable waiting time dave _{i of the} queue buffer i and the allowable waiting time dp of the input packet is calculated,
For each queue buffer i, calculate the fitness r _i which is a ratio of the path bandwidth w _i for deviance g _i,
Calculating a route selection probability p _i which is a ratio of the fitness r _i of the queue buffer i to the fitness Σr _i of all the queue buffers;
By arranging all the queue buffers in order and accumulating the route selection probability p _i , the route selection probability range for each queue buffer i is determined,
When a packet is input, it is also preferable to generate a random number, select a queue buffer that matches the path selection probability range including the random number value, and output the packet to the queue buffer.

According to another embodiment of the communication device of the present invention,
About route selection means,
The divergence degree g _i is calculated as follows using the target allowable standby time k _i · dave _i obtained by multiplying the average allowable standby time dave _i by the aggregation coefficient k _i : g _i = k _i · dave _i −dp
It is also preferable.

According to another embodiment of the communication device of the present invention,
The aggregation coefficient k _i is a moving average value of the aggregate packet size obtained by multiplying the past aggregation coefficient k _i ′ by the ratio of the target transmission size L to the actual transmission size L _i of the aggregate packet output from the queue buffer. K _i = max (1, k _i '· L / L _i ) calculated as follows:
max (x, y): It is also preferable to output the larger value of x or y.

According to another embodiment of the communication device of the present invention,
The fitness r _i is calculated using the divergence g _i as follows: r _i = w _i / log (1 + g _i ² )
= W _i / log (1+ (k _i · dave _i −dp) ² )
It is also preferable.

According to another embodiment of the communication device of the present invention,
A time stamp giving means for giving a time stamp to each input packet;
A permissible standby time calculating means for giving a permissible standby time dp that is a time difference between the time stamp and the current time for each packet;
It is also preferable that a packet with a time stamp and an allowable waiting time dp is input to the route selection means.

According to another embodiment of the communication device of the present invention,
The route selection means
A flow table in which the header information of the packet is registered in association with the identifier of the queue buffer to which the packet is to be input;
Determine whether the header information of the input packet matches the header information registered in the flow table,
If it is determined to be true, the packet is output to the registered queue buffer,
When it is determined to be false, the packet is selected from a plurality of queue buffers according to the allowable waiting time dp of the packet, and the identifier of the selected queue buffer and the header information of the input packet are associated with each other in the flow table. It is also preferable to register.

According to another embodiment of the communication device of the present invention,
The communication device is an edge device installed at the boundary between the core network and the access network,
The header information is preferably information based on protocol headers from layer 1 to layer 4.

According to the present invention, there is provided a system having the plurality of communication devices described above and a management device that manages a route bandwidth for each route in the network,
Management device, to each communication apparatus, transmits a new path band n _i for each path i,
Communication device, the actual measurement path band s _i of each queue buffer i is measured, the path bandwidth w _i to be set for each queue buffer i, is calculated by the following equation _{w i = n i × (n} i / s i )
It is characterized by that.

According to the present invention, in a program for causing a computer mounted on a communication device having a queue buffer different for each path to function in order to transfer an input packet to a different path,
The input packet is given an allowable waiting time dp,
For each queue buffer i, the path bandwidth w _i in the path output from the queue buffer i is set,
The queue buffer accommodates and outputs a plurality of packets in one aggregated packet when the allowable waiting time dp is reached for the accumulated packets,
A queue buffer having a narrower route bandwidth w _i among a plurality of queue buffers is characterized in that the computer is caused to function as route selection means for inputting a packet having a larger allowable waiting time dp.

According to the present invention, in a packet input / output method in a device having a queue buffer different for each path in order to transfer an input packet to a different path,
The input packet is given an allowable waiting time dp,
For each queue buffer i, the path bandwidth w _i in the path output from the queue buffer i is set,
The queue buffer accommodates and outputs a plurality of packets in one aggregated packet when the allowable waiting time dp is reached for the accumulated packets,
Among the plurality of queue buffer, the narrower the queue buffer path bandwidth w _i, characterized by inputting a large packet is allowed waiting time dp.

  According to the apparatus, method, and program of the present invention, the accommodation efficiency of aggregated packets for each path can be increased in consideration of the allowable waiting time. That is, according to the apparatus of the present invention, both (1) maximizing the network utilization rate based on route selection of each packet and (2) maximizing the allowable average packet length based on aggregation of a plurality of packets. The effect of can be obtained.

1 is a system configuration diagram for communicating between edge devices via a core network. FIG. It is a functional block diagram of the edge apparatus in this invention. It is explanatory drawing showing the basic function of the route selection part in this invention. . It is explanatory drawing showing the specific process of the route selection part in this invention. It is explanatory drawing showing control of the route band of the queue buffer from a management apparatus.

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

  FIG. 2 is a functional configuration diagram of the edge device according to the present invention.

  FIG. 2 shows a communication device 1 that outputs an aggregate packet containing a plurality of input packets. The communication device 1 is an edge device installed at the boundary between a core network and an access network, and transmits and receives aggregated packets to and from a partner edge device via the core network. 2 includes an input interface 10, a time stamp assigning unit 11, an allowable standby time calculating unit 12, a route selecting unit 13, a plurality of queue buffers 14, and an output interface 15. These functional components can be realized by executing a program that causes a computer installed in the apparatus to function. The processing flow of these functional components can also be understood as an input / output method in the apparatus.

  The input interface 10 is connected to the access network and inputs packets from the access network. These packets are output to the time stamp assigning unit 11.

[Time stamp giving unit 11]
The time stamp assigning unit 11 assigns the time immediately after input to each input packet (time stamp). Based on the time stamp for each packet, the allowable waiting time in the packet can be calculated for the subsequent processing. The packet is output to the allowable standby time calculation unit 12.

[Allowable standby time calculation unit 12]
The allowable standby time calculation unit 12 assigns an “allowable standby time” (allowable delay time) that is a time difference between the time stamp and the current time for each packet. The allowable waiting time means a time during which it is allowed to stay in the queue buffer in the edge device.
Allowable standby time dp = current time−time stamp The packet to which the allowable standby time dp is assigned is output to the route selection unit 13.

[Route selection unit 13]
The edge device 1 establishes a plurality of routes to the same destination edge device 1 using a route control protocol such as OSPF. A packet added with a time stamp and an allowable waiting time dp is input to the route selection unit 13, and the packet is distributed to any queue buffer 14. The subsequent queue buffer 14 is provided for each path.

  FIG. 3 is an explanatory diagram showing the basic function of the route selection unit in the present invention.

  The route selection unit 13 has a flow table in which a route (route identifier, that is, a queue buffer identifier) is stored in association with packet header information. The flow table specifies one destination edge device as a destination for a packet and registers a route to which the packet is to be transferred from among a plurality of routes established for the destination edge device. . Here, the header information may be “flow tuple”. A flow tuple is information based on a protocol header from layer 1 (for example, a physical port of a switch) to layer 4 (for example, a TCP / IP port number). For example, when the present invention is applied to an OpenFlow switch, there are about 14 types of flow tuples, and the flow table can be set without being based on a specific layer.

  The route selection unit 13 first refers to the flow table and determines whether or not the header information of the input packet is registered. If already registered (YES), the packet is output to the route registered in the flow table, that is, to the registered queue buffer 14. For packet flows, once established paths are used continuously.

On the other hand, when the header information of the input packet is not registered in the flow table (NO), the route selection unit 13 selects the relevant route from a plurality of routes established with respect to the destination partner edge device. Select the route that the packet should be forwarded to. At this time, the queue buffer is selected according to the allowable waiting time dp of the packet.

According to FIG. 3, the queue buffer 14 has at least “path bandwidth w _i ” in the path output from the queue buffer i for each queue buffer i. Here, the queue buffer _i having a narrower path bandwidth w _i requires a longer buffer waiting time for packet aggregation. When packets with a short allowable waiting time are stored in such a queue buffer i, the packets must be collected and transmitted before a sufficient number of packets accumulate in the queue buffer. In this case, the packet accommodation efficiency decreases. For this purpose, the path selection unit 13 in the edge device 1 of the present invention performs control so that a queue buffer i having a narrower path bandwidth w _i among a plurality of queue buffers receives a packet having a larger allowable waiting time dp.

  FIG. 4 is an explanatory diagram showing specific processing of the route selection unit in the present invention.

FIG. 4 shows a specific process for the route selection unit 13 to input a packet having a longer allowable waiting time dp as the queue buffer i having a smaller route bandwidth w _i . First, the queue buffer 14 stores “average allowable waiting time dave _i ” for a plurality of packets input in the past for each queue buffer i. Then, the route selection unit 13 executes the following steps.

(S1) For each queue buffer i, a deviation g _i between the average allowable waiting time dave _{i of the} queue buffer i and the allowable waiting time dp of the input packet is calculated. The degree of divergence g _i is calculated as follows using the aggregation coefficient k _i .
g _i = k _i · dave _i −dp
g _i : Degree of deviation of queue buffer i
k _i : aggregation factor of queue buffer i
dave _i : Average allowable waiting time of queue buffer i
dp: Allowable waiting time of the packet

The aggregation coefficient k _i is used to derive the target allowable waiting time k _i · dave _i of the packet waiting in the queue buffer by increasing the packet aggregation efficiency. The aggregation coefficient k _i may be a static set value or may be a value linked to the moving average value of the aggregate packet size as described below. The aggregation coefficient k _i is a moving average value of the aggregate packet size obtained by multiplying the past aggregation coefficient k _i ′ by the ratio of the target transmission size L to the actual transmission size L _i of the aggregate packet output from the queue buffer. Is calculated as follows.
k _i = max (1, k _i '· L / L _i ) (k _i ≧ 1)
max (x, y): Outputs the larger value of x or y
k _i ': Past aggregation coefficient of queue buffer i
L: Target transmission size
L _i : Actual transmission size of queue buffer i Here, in order to make it easy to distribute packets having a larger target allowable waiting time k _i · dave _i with respect to the average allowable waiting time dave _i of the path i, the aggregation coefficient k _i is multiplying.

According to the above formula, if the aggregate packet size is the target transmission size L, it is assumed that the aggregation efficiency is maximum. When the measured transmission size L _i is smaller than the target transmission size L, the aggregation coefficient is increased so as to approach the target transmission size L. Further, the fact that the actually measured transmission size L _i is smaller than the target transmission size L means that a small number of packets are aggregated and transmitted, that is, the allowable waiting time of the route is not sufficiently large. To do. Therefore, by increasing the aggregation coefficient k _i , packets having a longer allowable waiting time can be distributed to the route i, and the aggregation efficiency can be recovered.

(S2) Next, for each queue buffer i, and calculates the matching degree r _i which is a ratio of the path bandwidth w _i with respect to discrepancy g _i. The fitness r _i is calculated using the divergence g _i as follows: r _i = w _i / log (1 + g _i ² )
= W _i / log (1+ (k _i · dave _i −dp) ² )
r _i : Goodness of fit
g _i : Degree of divergence
w _i : Path bandwidth The closer the allowable standby time dp of the packet is to the target allowable standby time k _i · dave _i that is larger than the average allowable standby time dave _i , the greater the fitness r _i . The log used for calculating the fitness r _i is for increasing the route selection probability p _i with higher sensitivity as the allowable waiting time dp of the packet is closer to the target allowable waiting time k _i · dave _i. is there.

(S3) Next, to calculate the route selection probability p _i as a percentage of the fitness r _i of the queue buffer i for fit? R _i of all queue buffer.
p _i = r _i / Σr _i
The greater the fitness r _i , the greater the selection probability p _i of the route i. As a result, the packet is easily selected as the route i.

(S4) Then, by accumulating route selection probability p _i side by side all the queue buffer in order to determine the path selection probability range for each queue buffer i. The cumulative sum of the route selection probabilities p _i is 1. According to FIG. 4, the queue buffers are allocated as follows, for example.
Σ _{k = 0} ^{k = i−1} p _k ≦ route selection probability range for each queue buffer i <Σ _{j = 0} ^{j = i} p _j
The queue buffer 1 has a route selection probability p _i = 0.2 and is assigned a route selection probability range 0 (or more) to 0.2.
The queue buffer 2 has a route selection probability p _i = 0.3 and is assigned a route selection probability range of 0.2 (or more) to 0.5.
Eventually, the route selection probability range 0.95 (or higher) to 1 is assigned to the queue buffer n if the route selection probability p _i = 0.05.

(S5) When a packet is input, a uniform random number a in the range of 0 to 1 is generated. Then, the random number a selects a queue buffer assigned to the following route selection establishment range.
Σ _{k = 0} ^{k = i-1} p _k ≦ a <Σ _{j = 0} ^{j = i} p _j
According to FIG. 4, for example, it is assumed that a random value 0.15 is generated. At this time, the queue buffer 1 that matches the route selection probability range 0 to 0.2 including the random value is selected. Although not shown, for example, when 0.4 out of the lances is generated, the queue buffer 2 that matches the route selection probability range 0.2 to 0.5 including the random value is selected.

(S6) The route selection unit 13 outputs the packet to the selected queue buffer i.

(S7) After the queue buffer is selected for the packet, the header information of the packet and the queue buffer identifier are associated with each other and registered in the flow table. Thereafter, all the flows of the packet are output to the same queue buffer according to the flow table.

[Queue buffer 14]
The queue buffer 14 is a FIFO (First-In First-Out) buffer, and is provided for each path. The queue buffer 14 accommodates and outputs a plurality of packets in one aggregated packet in order from the output side when any of the accumulated packets reaches the allowable waiting time dp.

Here, it is preferable that the queue buffer i for each path sequentially recalculates the average allowable waiting time d _i , the measured transmission size Li, and the measured path bandwidth s _i . As a specific calculation method, the following moving average may be used.

Calculation method of average allowable waiting time d _i d _i = α · dp + (1−α) d _i ′
d _i ': Existing allowable average waiting time alpha: coefficient dp: According to this calculation method allowable waiting time of the packet, for each queue buffer, so that the allowable average latency d _i are calculated successively.

Method for calculating actual transmission size L _i of aggregated packets L _i = β · Li + (1−β) Li ′
L _i ′: Actual transmission size of existing aggregate packet β: Coefficient According to this calculation method, the actual transmission size L _i of the aggregate packet is sequentially calculated for each queue buffer.

Note that the measured path bandwidth s _i is calculated by dividing the size of the aggregate packet by the transmission time (transmission end time−transmission start time).

According to the above-described embodiments of FIGS. 2 to 4, the edge device moves a packet having a large allowable waiting time dp to a path having a relatively slow bandwidth, and conversely, a packet having a small allowable waiting time dp is relatively fast. To a route with a bandwidth of. In addition, the allowable average packet length of the packets transferred to each path can be maximized. Thus, a route having a high speed band can transfer more packets than a route having a low speed band. Further, after being registered in the flow table, all the packets constituting the flow are transferred to the same route. As a result, the packet order is not reversed with respect to the flow.

  FIG. 5 is an explanatory diagram showing control of the path bandwidth of the queue buffer from the management apparatus.

According to FIG. 5, each edge device 1 transmits the actually measured route bandwidth s _i for each route to the management device 2. The management device 2 receives the actually measured route bandwidth s _i for each route from the multiple edge devices 1 connected to the core network. For example, an OSPF protocol can be used to detect the topology (path) of the core network. Further, for example, a linear programming method can be used for the optimization calculation of the actually measured path bandwidth from the utilization status of all the lines in the core network.

In addition, the management device 2 transmits a new route bandwidth _ni for each route _i to each edge device 1 at a fixed period Ic, and each edge device 1 manages the measured route bandwidth s _i for each route. 2 to send. The fixed period Ic is sufficiently longer than the average allowable waiting time d _i and the update period Ir of the actually measured transmission size Li in the edge device 1.

In contrast, each edge device 1, the measured path band s _i of each queue buffer i is measured, the path bandwidth w _i to be set for each queue buffer i, is calculated by the following equation.
w _i = n _i × (n _i / s _i )

  As described above in detail, according to the apparatus, method, and program of the present invention, it is possible to increase the accommodation efficiency of aggregated packets for each path in consideration of the allowable waiting time. That is, according to the apparatus of the present invention, (1) maximization of the network utilization rate based on route selection of each packet and (2) maximization of the allowable average packet length based on aggregation of a plurality of packets are obtained. It is done.

  Various changes, modifications, and omissions of the above-described various embodiments of the present invention can be easily made by those skilled in the art. The above description is merely an example, and is not intended to be restrictive. The invention is limited only as defined in the following claims and the equivalents thereto.

DESCRIPTION OF SYMBOLS 1 Edge device, communication apparatus 10 Input interface 11 Time stamp provision part 12 Permissible waiting time calculation part 13 Path selection part 14 Queue buffer group 15 Output interface 2 Management apparatus

Claims (11)

In a communication apparatus having a different queue buffer for each path in order to transfer an input packet to a different path,
The input packet is given an allowable waiting time dp,
For each queue buffer i, a path bandwidth w _i in a path output from the queue buffer i is set,
The queue buffer accommodates and outputs a plurality of packets in one aggregated packet when the allowable waiting time dp is reached for the accumulated packets,
A communication apparatus comprising path selection means for inputting a packet having a larger allowable waiting time dp to a queue buffer having a narrower path bandwidth w _i among a plurality of queue buffers. For each queue buffer i, an average allowable waiting time dave _i for a plurality of packets input in the past is further recorded,
The route selection means includes
For each queue buffer i, a deviation g _i between the average allowable waiting time dave _{i of the} queue buffer i and the allowable waiting time dp of the input packet is calculated,
Wherein each queue buffer i, calculate the fitness r _i which is a ratio of the path bandwidth w _i with respect to the deviation degree g _i,
Calculating a route selection probability p _i which is a ratio of the fitness r _i of the queue buffer i to the fitness Σr _i of all the queue buffers;
By accumulating the route selection probability p _i arranged all queue buffer in order to determine the path selection probability range of the respective queue buffer i,
The random number is generated when a packet is input, a queue buffer that matches a path selection probability range including the random value is selected, and the packet is output to the queue buffer. The communication device described. About the route selection means,
The divergence degree g _i is calculated as follows using the target allowable standby time k _i · dave _i obtained by multiplying the average allowable standby time dave _i by the aggregation coefficient k _i : g _i = k _i · dave _i − dp
The communication device according to claim 2. The aggregation coefficient ki is a moving average value of the aggregate packet size obtained by multiplying the past aggregation coefficient k _i ′ by the ratio of the target transmission size L to the actual transmission size L _i of the aggregate packet output from the queue buffer. And k _i = max (1, k _i '· L / L _i ) calculated as follows:
The communication apparatus according to claim 3, wherein max (x, y): x or y, whichever is larger, is output. The fitness r _i is calculated using the divergence g _i as follows: r _i = w _i / log (1 + g _i ² )
= W _i / log (1+ (k _i · dave _i −dp) ² )
The communication apparatus according to claim 3 or 4, wherein A time stamp giving means for giving a time stamp to each input packet;
An allowable standby time calculating means for giving an allowable standby time dp that is a time difference between the time stamp and the current time for each packet;
6. The communication apparatus according to claim 1, wherein a packet added with the time stamp and the allowable waiting time dp is input to the route selection unit. The route selection means includes
A flow table in which the header information of the packet is registered in association with the identifier of the queue buffer to which the packet is to be input;
Determine whether the header information of the input packet matches the header information registered in the flow table,
If it is determined to be true, the packet is output to the registered queue buffer,
If it is determined to be false, the flow table is selected from a plurality of queue buffers according to the allowable waiting time dp of the packet, and the identifier of the selected queue buffer and the header information of the input packet are associated with each other. The communication apparatus according to any one of claims 1 to 6, wherein the communication apparatus is registered. The communication device is an edge device installed at the boundary between the core network and the access network,
The communication apparatus according to claim 7, wherein the header information is information based on a protocol header from layer 1 to layer 4. A system comprising a plurality of communication devices according to any one of claims 1 to 8, and a management device that manages a route bandwidth for each route in the network,
The management device, to each communication apparatus, transmits a new path band n _i for each path i,
The communication device, the actual measurement path band s _i of each queue buffer i is measured, the path bandwidth w _i to be set for each queue buffer i, is calculated by the following equation _{w i = n i × (n} i / s _i )
A system characterized by that. In a program for causing a computer mounted on a communication device having a different queue buffer for each route to function in order to transfer an input packet to a different route,
The input packet is given an allowable waiting time dp,
For each queue buffer i, a path bandwidth w _i in a path output from the queue buffer i is set,
The queue buffer accommodates and outputs a plurality of packets in one aggregated packet when the allowable waiting time dp is reached for the accumulated packets,
A program that causes a computer to function as route selection means for inputting a packet having a larger allowable waiting time dp as a queue buffer having a narrower route bandwidth w _i among a plurality of queue buffers. In a packet input / output method in a device having a different queue buffer for each path in order to transfer an input packet to a different path,
The input packet is given an allowable waiting time dp,
For each queue buffer i, a path bandwidth w _i in a path output from the queue buffer i is set,
The queue buffer accommodates and outputs a plurality of packets in one aggregated packet when the allowable waiting time dp is reached for the accumulated packets,
A packet input / output method characterized by inputting a packet having a larger allowable waiting time dp to a queue buffer having a narrower path bandwidth w _i among a plurality of queue buffers.

Images