JP4391346B2 - COMMUNICATION CONTROL METHOD, COMMUNICATION CONTROL DEVICE, CONTROL PROGRAM, AND RECORDING MEDIUM - Google Patents

Description

  The present invention relates to a communication control method for fair distribution of surplus bandwidth and buffer capacity in a communication control method using a token bucket model, a communication control device using the method, a control program, and a recording medium.

  In recent years, in connection services for communication networks, high-speed connection services can be provided at low cost due to the widespread use of FTTH (Fiber To The Home).

  For example, VoIP (Voice over IP) and Internet telephone which digitize and compress voice data and exchange or transmit the data on the Internet or IP network have been put into practical use. Since VoIP can transmit data packets and voice data mixedly on a single network, it can save both network construction costs and operation costs. In addition, communication services that guarantee bandwidth available to users are becoming widespread.

  Data traffic flowing through such a communication network needs to be subjected to QoS (Quality of Service) control according to the application used.

  With QoS control, SLA (Service Level Agreement) is enforced through bandwidth adjustment. SLA defines the assurance criteria that a network must meet for a specified user or service type. These criteria include guaranteed bandwidth, packet loss, and the like.

  As the guaranteed bandwidth, the minimum, average, and peak guaranteed values of the usable bandwidth can be designated. Packet loss indicates the number or percentage of packets that have not been received after transmission, or packets that have received errors.

  As a technical method for performing the QoS control, there is a scheduling method in which different queues are configured for each flow, and packets transmitted from the queues are extracted one by one by a scheduler and scheduled.

  Also, each communication control device (hereinafter referred to as a router) in the network classifies flows into flows for each user and service type, assigns a bandwidth and buffer capacity to each flow, and sets the maximum flow rate of the flow. Restricting policing methods are also used.

FIG. 5 is a diagram illustrating one aspect of the scheduling method.
FIG. 5 includes a scheduler 13 having a plurality of information sources 25-1 to 25 -n, queues 12-1 to 12 -n, flows 32-1 to 32 -n, and one outgoing link 17.

  As shown in FIG. 5, the scheduling method is a method in which packets are transmitted from a plurality of information sources 25-1 to 25-n, and queues 12-1 to 12-n are respectively transmitted through flows 32-1 to 32-n. to go into. The scheduler 13 can retrieve only one packet from the queues 12-1 to 12-n at a time.

  When there are a plurality of flows, the scheduler 13 needs to wait for the packets in a plurality of queues, take out the packets one by one from the queue in accordance with a predetermined method, and send them to the outgoing link. This method of selecting the transmission order is called a scheduling method.

  Typical scheduling methods include FIFO (First In First Out) scheduling for sending packets to the outgoing link in the order of arrival, insufficient round-robin scheduling for handling variable-length packets, weighted fair scheduling for fair handling of flows, etc. There is.

  There is also priority scheduling in which an application determines a priority (priority) for each flow, adds the flow priority to all packets of the flow, and processes the packets according to the priority determined by the priority.

  On the other hand, the number of users accommodated in a single router is increasing for reasons such as cost reduction, and it is not uncommon to accommodate thousands to tens of thousands of users. Also, the line speed has been rapidly increased in recent years.

  However, it is difficult in terms of processing capability of the router to individually configure queues for a large number of users on such a high-speed line and perform packet scheduling. Even if it is realized by dedicated hardware, the design circuit becomes complicated and the cost is very high.

  In addition, the scheduling method has a problem in that the utilization efficiency of the packet buffer capacity becomes inefficient. Assume that the buffer capacity of the router is 1 megabyte. If queues are individually created for 1000 users and the maximum queue length of each queue is set, 1 megabyte ÷ 1000 users = 1 buffer capacity is allocated.

  However, even if there are 1000 users who have subscribed to the service, the number of users who are actually making a flow at a certain time is not so large. For example, if the number of users contracting for the service is 400, 600 users do not actually flow.

  Therefore, if the maximum queue length is set as described above, the packet capacity will be discarded or delayed even though the buffer capacity is practically scarcely used, and the efficiency of using the buffer capacity becomes very poor. .

  Furthermore, as disclosed in Patent Document 2, it is difficult to efficiently allocate buffer capacity by using one scheduling method among a plurality of scheduling methods, and the implementation becomes extremely complicated.

  On the other hand, instead of the scheduling method, a communication control method capable of limiting the bandwidth and the burst amount for each flow, such as a policing method, is also provided.

  In the policing method, the bandwidth and the allowable burst amount are verified for each flow. In other words, the policing method always measures the flow using a flow meter called a policer. When the bandwidth and allowable burst amount are tested and determined to be nonconforming, that is, when the flow exceeds the bandwidth value set for the policer, or when the flow exceeds the allowable burst amount set for the policer The packet is discarded or delayed.

As a result of the verification of the bandwidth and the allowable burst amount, if it is determined to be compatible, that is, the flow does not exceed the bandwidth value set in the policer, and the flow exceeds the allowable burst amount set in the policer. If not, the packet is a method of entering the next stage of processing.

  As one of the policing methods, there is a policing method using a token bucket model.

FIG. 6 is a diagram schematically illustrating a polishing method using a token bucket model.
FIG. 6 includes a token 26, a token bucket 27, a packet 28, a policer 8, and a queue 12.

  As shown in FIG. 6, the policing method using the token bucket model uses a virtual flow rate representing a traffic transmission right called a token 26. A virtual token bucket 27 for storing the token 26 is prepared for the flow whose flow rate is to be limited. The token bucket 27 is replenished with the token 26 at a constant rate equal to the flow limit flow rate, that is, at a speed at which the token 26 is supplied to the token bucket 27.

  When the packet 28 arrives at the policer 8, the policer 8 takes out the token 26 proportional to the data amount of the packet from the token bucket 27. For example, when the packet 28 arrives at the policer 8 and the packet 28 requires the token 26 for 64 bytes, the token 26 for 64 bytes is extracted from the token bucket 27.

  The packet 28 obtains the traffic transmission right by extracting the required amount of tokens 26, and the packet 28 is transmitted and temporarily stored in the queue 12. When the token 26 required for the packet 28 is not supplied from the token bucket 27, the packet 28 is subjected to processing such as discarding and delaying. By doing so, flow flow restriction is realized.

However, the above-described policer performs bandwidth limitation for each of a plurality of flows as disclosed in Patent Document 1. In this case, if there is a flow in which no packet is transmitted, the surplus bandwidth cannot be used by another flow, and the buffer capacity cannot be efficiently allocated. Furthermore, it is extremely difficult to implement the distribution of surplus bandwidth to each flow evenly.
JP-A-6-178373 JP-T-2001-519120

  However, the above-described policing method using the token bucket model has a problem in that each flow is verified and only a function of discarding or delaying is provided, and a bandwidth and a buffer capacity cannot be allocated fairly. It was.

  The present invention has been made in view of the above-described problems. A communication control method capable of performing bandwidth limitation on each flow and fairly distributing the bandwidth and the buffer capacity to each flow, and its An object of the present invention is to provide a communication control device, a control program, and a recording medium using the method.

A first aspect of the communication control method of the present invention is a communication control method for performing transmission by performing bandwidth control on input packet data by policing processing ,
Classifying the input packet data into flows for at least one user;
A policing step for allocating all the bandwidths allocated for the transmission to the flows to be transmitted in parallel to the flows to be actually transmitted ,
The flow is assigned a weight according to the bandwidth ratio,
The policing step includes:
When the weight of the flow of the flow number i that is actually communicating is W _i , the total value of the weights of the entire flow that is currently communicating is A, and the packet data length is L,
The theoretical expected value (TAT value) of the arrival time of the next packet of the flow number i is calculated from the formula TAT _i = TAT _i + L × A / W _i
By performing the policing process using the predicted value,
This is a communication control method characterized by allocating a bandwidth based on the ratio of W _i / A to a flow that is actually communicating.

According to a second aspect of the communication control method of the present invention, the transmission rate of data transmitted by at least one user is varied when allocating the flow to be actually transmitted at least for each user. It is.

A third aspect of the communication control method of the present invention is a communication control method characterized in that the packet capacity to be allocated is varied at least for each user.

A first aspect of the communication control apparatus of the present invention is a communication control apparatus that performs transmission by performing bandwidth control on input packet data by policing processing ,
A packet classification mechanism for classifying the input packet data into flows for at least one user;
A policing processing unit that allocates the entire bandwidth allocated for transmission to a plurality of flows that can be transmitted in parallel, to the flow to be actually transmitted ,
The flow is assigned a weight according to the bandwidth ratio,
The polishing processing unit
When the weight of the flow of the flow number i that is actually communicating is W _i , the total value of the weights of the entire flow that is currently communicating is A, and the packet data length is L,
The theoretical expected value (TAT value) of the arrival time of the next packet of the flow number i is calculated from the formula TAT _i = TAT _i + L × A / W _i
By performing the policing process using the predicted value,
The communication control apparatus is characterized in that a bandwidth is allocated based on a ratio of W _i / A to a flow that is actually communicating .

According to a second aspect of the communication control apparatus of the present invention, at least for each user, when allocating to a flow to be actually transmitted, a transmission rate of data transmitted by at least one user is varied. It is.

A third aspect of the communication control apparatus according to the present invention is a communication control apparatus characterized in that a packet capacity to be allocated is varied at least for each user.

According to a first aspect of the control program of the present invention, there is provided a control program for controlling, by a computer, a communication control device that performs transmission by performing bandwidth control by policing processing on input packet data.
Classifying the input packet data into at least one user flow;
A policing process step of allocating the entire bandwidth allocated for the transmission to a plurality of flows that can be transmitted in parallel to the flow to be actually transmitted ;
The flow is assigned a weight according to the bandwidth ratio ,
The policing processing step includes:
When the weight of the flow of the flow number i that is actually communicating is W _i , the total value of the weights of the entire flow that is currently communicating is A, and the packet data length is L,
The theoretical expected value (TAT value) of the arrival time of the next packet of the flow number i is calculated from the formula TAT _i = TAT _i + L × A / W _i
By causing the policing processing step to be executed using the predicted value,
This is a control program characterized by causing a flow in actual communication to perform bandwidth allocation based on the ratio of W _i / A.

According to a second aspect of the control program of the present invention, there is provided a control program characterized by varying the transmission rate of data transmitted by at least one user when allocating to a flow to be actually transmitted at least for each user. .

A third aspect of the control program of the present invention is a control program characterized in that the packet capacity to be allocated is varied at least for each user.

  A first aspect of the recording medium of the present invention is a computer-readable recording medium for recording the control program described above.

  According to the communication control method of the present invention, a bandwidth and a buffer capacity can be allocated to each user's flow in a fair manner by the simple communication control method shown in the present invention without performing a complicated scheduling method.

  Further, it is possible to allocate a buffer capacity of a packet in consideration of only a user who is communicating, which is difficult with a scheduling method. For example, if a device that accommodates 1000 users and the buffer capacity is 1 megabyte, and there are only 10 users communicating at a certain point in time, the scheduling method uses 1 kilobyte for each user. A buffer capacity of a capacity is allocated.

  However, when the communication control method of the present invention is used, when there are only 10 users who are communicating at a certain time, it is possible to allocate a buffer capacity of 100 kilobytes.

  Embodiments of the present invention will be described below in detail with reference to the drawings.

FIG. 1 is a diagram illustrating a structure of a communication control apparatus including a policer.
As shown in FIG. 1, the router 1 includes input / output ports 2-1 to 2-n, a switch fabric 3, and a network processor 33. The network processor 33 includes a traffic adjustment mechanism 6 and a traffic control mechanism 15.

  The router 1 includes a CPU (not shown) for controlling the entire router 1 and a ROM, RAM, etc. (not shown) for storing a control program or control data for communication control. Control is performed by the CPU based on a control program in the ROM.

    The switch fabric 3 is a mechanism for switching a packet received at each input port to an appropriate destination port.

  The traffic adjustment mechanism 6 includes a packet classification mechanism 5 that classifies packets for each user and each service type, and a packet adjustment mechanism 7 that discards and delays packets sent from the packet classification mechanism 5. The packet classification mechanism 5 includes a classifier, for example, and uses a CAM (Contents Addressable Memory).

  That is, the traffic adjustment mechanism 6 checks whether or not the arriving packet conforms to the contracted flow characteristics and performs a predetermined process specified in advance. The designated predetermined process includes a process for discarding a packet and a process for delaying a packet.

  The packet adjustment mechanism 7 discards the packet sent from the packet classification mechanism 5 and discards the packet determined to be discarded by the policer 8 and the policer 8 that determines whether or not the delay processing and the delay processing may be performed. A processing unit (dropper) 9 is provided.

  The traffic control mechanism 15 includes a queue manager 11 that sends packets to the queue 12, a queue 12 that contains the sent packets, and a scheduler 13 that selects packets to be output to the outgoing link 17. A rectangular dotted line surrounding the queue 12 in the traffic control mechanism 15 represents the buffer capacity 16. That is, the queue manager 11 and the scheduler 13 are mechanisms for controlling packets entering and leaving the queue 12 at the entrance and exit of each queue 12.

  That is, the traffic control mechanism 15 controls the selection and transmission order of packets to be sent to the outgoing link 17 so as to satisfy the requirements such as the output rate and delay time of each flow.

  First, when a packet is input, the switch fabric 3 having a switch function refers to the destination IP address information in the IP header portion of the packet input from each input port, for example, and switches to a predetermined destination port. I do. The packet subjected to the switching process is sent to the packet classification mechanism 5.

  Next, the switched packets are classified by the packet classification mechanism 5 into individual flows for each user and each service type. The packet classification mechanism 5 performs classification with reference to, for example, a transmission / reception IP address in a packet IP header and a transmission / reception port in a TCP / UDP header. At this time, an internal header indicating the flow number is added to the header portion of the packet and output to the policer at the next stage.

  Furthermore, the policer 8 tests the bandwidth and the allowable burst amount for each flow. That is, the flow is constantly measured using a flow meter called a policer 8. As a result of the test, if the flow is determined to be incompatible, that is, if the flow exceeds the bandwidth value set for policer 8, or if the allowable burst amount set for policer 8 is exceeded, Time processing is performed. For example, the packet is discarded and transferred to the discard processing unit 9.

  As a result of the test, if it is determined that the flow is suitable, that is, if the flow does not exceed the bandwidth value set in the policer and does not exceed the allowable burst amount set in the policer, the packet is transmitted to the traffic control mechanism 15. Forwarded to

  The queue manager 11 selects a packet accommodated in the queue 12. In the queue 12, the packets for the allowable burst of the policer in the previous stage are retained and delayed to the output port speed. Then, the scheduler 13 waits for the packets in one or a plurality of queues, and sequentially extracts the packets one by one from the queue according to a predetermined method. The packet selected as one according to a predetermined method is output to the outgoing link 17 by the scheduler 13 and transmitted.

  FIG. 2 is a diagram for schematically explaining the structure of the policer.

  As shown in FIG. 2, the policer 8 includes a policing processing unit 18, a flow table 19, a rate storage unit 20, an active counter unit 21, a burst storage unit 22, a flow state storage unit 23, and a timer unit 24.

  The rate storage unit 20 stores a total value R (bps) of the rates obtained by adding the rates in each flow. The rate represents the speed at which tokens are supplied to the token bucket. For example, it may be equal to the value of the output port rate. The burst storage unit 22 stores an allowable burst value B for policing. The allowable burst value indicates a value that limits the length of a packet sequence that is continuously transmitted. For example, the allowable burst value may be equal to the buffer capacity of the packet assigned to the output port.

  The flow state storage unit 23 stores the value of the check interval T for changing the state of each flow from the Active state to the Inactive state. The timer unit 24 stores the current time t. The active counter unit 21 stores a total value A of weights obtained by summing up the weight values of the flows that are currently communicating, that is, in the active state.

FIG. 3 is a diagram illustrating a configuration example of a flow table.
As shown in FIG. 3, the flow table is a table having a flow number i (i is a natural number), a weight W _i , a TAT _i value (Theoretical Arrival Time), and a flag F _i for managing a communication state. I indicated by each subscript indicates a weight, a TAT value, and a flag for each flow number. The weight W _{i of the} flow table represents the ratio of the bandwidth in each flow. For example, when there is a setting request for the bandwidths of 2 Mbps, 1 Mbps, and 4 Mbps for each of the flow numbers 1, 2, and 3, each of the flows 1, 2, and 3 has a weight of 2: 1: 4.

The TAT _i value in the flow table indicates the theoretical expected value of the arrival time of the next packet with flow number i. The Active registered in the flag F _i of flow table shows a state in which communication is performed, In addition, the Inactive, communication indicates a state not carried out.

  Next, a policing method for distributing the surplus bandwidth fairly will be described in detail.

FIG. 4 is a flowchart showing an operation of the policing processing unit of the policer shown in FIG.
The flowchart of FIG. 4 is based on a general token bucket algorithm. However, the fundamental difference from the general token bucket algorithm is that the sum A of the respective weights W _i of all flows currently in the active state is obtained. The point is that the bandwidth is allocated based on the ratio of the weight W _i of the flow in the active state that is stored in the active counter unit 21.

Specifically, when the packet length is added to TAT _i stored in the flow table 19, the correction packet length L1 represented by L1 = L × A / W _i is added, where L is the packet length. . In other words, even if the packet length is the same, a larger bandwidth is allocated to a packet of a flow having a higher weight among all flows that are currently in the active state.

More specifically, the policing rate of each flow is W _i / A times the total rate R, and bandwidth allocation is performed fairly according to the weight. Further, the burst value allowed for each flow is also W _i / A of the total allowable burst value B, and the buffer capacity can be allocated fairly according to the weight.

  First, the policing method makes a difference between the current time t and the time t ′ at which the Active state of each flow was checked last time. Then, it is determined whether or not the difference value is larger than the check interval T registered in the flow state storage unit (S101).

  At this time, if the difference value between the current time t and the time t ′ at which the active state of each flow was previously checked is greater than the value of the check interval T registered in the flow state storage unit (S101; YES), the flow 1 is assigned to the number i (S102). This substitution process means that the process is started from 1 for all the flow numbers i.

Next, it is determined whether the conditional expression TAT _i <R × t is satisfied (S103). That is, it is determined whether the token bucket is filled with tokens. Here, R represents a rate (speed) at which a traffic transmission right called a token is supplied to the token bucket.

When the conditional expression TAT _i <R × t is satisfied, that is, when the token bucket is satisfied with the token (S103; YES), the value calculated by R × t is substituted for TAT _i (S104). . The purpose of this assignment process is to cope with the fact that the bit width of the memory holding the current time t is finite and the value of t returns to 0 after a sufficiently long time. That is, the purpose is to allow the following calculation to be performed correctly by keeping the difference between TAT _i and the current time t within a certain range.

When the conditional expression TAT _i <R × t is not satisfied, that is, when the token bucket is not satisfied with the token (S103; NO), the value calculated by R × t is substituted for TAT _i. The process (S104) is not performed, and the next process is transferred to S107.

Next, it is determined whether or not the state of the flag F _{i in} the predetermined flow number i for managing the communication state is the active state (S105). State of the flag at a predetermined flow number i is an Active state (S105; YES) If the flag _{F i} to the Inactive state, reduces the active counter value by the weight _{W i} of the flow i (S106). As a result, a flow in another active state can be used in a band not used by the flow i.

The state of the flag _{F i} at a given flow number i is not Active state (S105; NO) case, the flag _{F i} to manage the communication and Inactive state, an active counter value by the weight _{W i} of the flow i decrease The process to be performed (S106) is not performed, and the next process is transferred to S107.

  Next, by adding 1 to the current flow number i, the process proceeds to the next flow number i (S107). For example, if the flow number i is 3 in the processing from S101 to S106, the flow number i becomes 4 in S107.

  Further, it is determined whether or not the conditional expression (flow number i) ≦ (last flow number i) (S108). The last flow number refers to the last flow number among the numbers added to the header part of the packet classified for each user and each service type by the packet classification mechanism. In this conditional expression, when the value obtained by adding 1 to the flow number i in S107 is equal to or smaller than the last flow number (S108; YES), the process returns to S103 and the next flow number is processed.

  If the value obtained by adding 1 to the flow number i in S107 exceeds the last flow number (S108; NO), the time t ′ at which the previous active state was checked is updated to the current time t (S109). Next, the time interval for checking the Active state is started.

  If t−t ′> T is not satisfied in S101 (S101; NO), that is, the difference value between the current time t and the time t ′ at which the Active state of each flow was checked last time is registered in the flow state storage unit. If it is not equal to the check interval T, the process proceeds to S110.

Subsequently, it is determined whether the packet has arrived at the policer (S110). Packet has arrived to the policer (S110; YES) If, furthermore, the flag _{F i} to manage the communication state determines whether the Inactive state (S 111).

  If the packet has not arrived at the policer (S110; NO), the process returns to S101, and the current time t is differentiated from the time t ′ at which the Active state of each flow was checked last time. Then, it is determined whether or not the difference value is equal to the value of the check interval T registered in the flow state storage unit.

If the flag _{F i} to manage the communication state is Inactive state (S 111; YES), the addition of the weight _{W i} of the flow i in the active counter, the flag _{F i} to manage the communication state to the Active state (S112).

Further, it is determined whether or not the conditional expression TAT _i <R × t is satisfied, that is, whether or not the token bucket is filled with tokens (S113). When the conditional expression TAT _i <R × t is satisfied (S113; YES), that is, when the token bucket is satisfied with tokens, the value calculated by R × t + L × A / W _i is substituted for TAT _i (S114). Then, the process proceeds to the transmission process (S118).

Here, L is shows a packet length, A is shows the total value which is the sum of the weights W _i for each flow number. That, L × value calculated by the A / W _i indicates a value obtained by doubling the packet length L by the ratio value and the weight W _i of the active counter.

When the conditional expression TAT _i <R × t is not satisfied, that is, when the token bucket is not satisfied with the token (S113; NO), the conditional expression TAT _i + L × A / W _i −R × t <B is further satisfied. It is determined whether or not (S115). Here, B represents a burst allowable value. This conditional expression determines whether or not there is an amount of tokens necessary to be supplied to the packet, which is the amount of tokens currently present in the token bucket.

When the conditional expression TAT _i + L × A / W _i −R × t <B is not satisfied (S115; NO), that is, the amount of tokens currently existing in the token bucket is only an amount necessary to be supplied to the packet. If the token amount does not exist, the packet is discarded (S116). Alternatively, once the packet is stored in a predetermined buffer capacity, etc., it can be set to be determined to be suitable when the remaining amount of tokens in the token bucket exceeds the required amount for the packet. It is.

When the conditional expression TAT _i + L × A / W _i −R × t <B is satisfied (S115; YES), that is, the amount of tokens currently existing in the token bucket and only the amount necessary to be supplied to the packet If the amount of the token is present, assigns the value calculated in TAT _i by _{TAT i + L × a / W} i to (S117), the packet is sent to the queue (S118).

FIG. 7 is a diagram illustrating a configuration of a network using a communication control device including a policer.
FIG. 7 includes users 30-1 to 30-5, LAN 29, router 1, and transmission paths 31 and 34. Although there is a network of several thousand users on one transmission line, here, it is assumed that there are five users for convenience. Currently, there are some transmission lines exceeding 10 Gbps. For convenience, the transmission line 31 is assumed to be 100 Mbps and the transmission line 34 is assumed to be 1 Gbps.
Packets received through the transmission line 34 at a maximum rate of 1 Gbps are adjusted to a maximum rate of 100 Mbps on the transmission line 31 by the policer of the present invention provided in the router 1.

  As shown in FIG. 7, when five users use a transmission path 31 of a maximum of 100 Mbps, a value obtained by dividing the maximum of 100 Mbps by the number of users of five is the fair bandwidth allocation realized by the policer at that time. In other words, in this case, the transmission path 31 used per person is assigned a bandwidth of 20 Mbps.

  Further, when the number of users is decreased from five to one, the value obtained by dividing the maximum 100 Mbps transmission path by the number of users that has been reduced is the fair bandwidth allocation at that time. For example, if the number of users is two, bandwidth allocation is 100 Mbps ÷ 2 = 50 Mbps.

  That is, the result of fair bandwidth allocation by the policer is as shown in Table 1. However, if it is not divisible, as in the case of using three people in Table 1, for example, a value obtained by rounding down the fraction is adopted. That is, 33 Mbps × 3 users. Or, assign as much as possible evenly. Alternatively, a virtual token bucket model that is divisible in advance is used. That is, for example, a virtual token bucket model for 34 Mbps × 3 users is set to 102 Mbps.

  In addition, when there is a user who is restricted depending on the content of communication contract, the response speed of the device, or the like, or when it is not fair, for example, user 1 has an upper limit of 10 Mbps, users 2 and 3 have an upper limit of 20 Mbps, When the users 4 and 5 are unlimited, it becomes as shown in Table 2. In other words, the limited users 1, 2 and 3 are constant at predetermined values. However, the users 4 and 5 divide the portion excluding the amount used by the users 1 to 3 fairly. Further, an arbitrary value can be assigned for each user regardless of the restriction.

The buffer capacity that the router 1 has on the port connected to the transmission line 31 is also assigned to each user fairly by the function of the policer as in the above-described band. For example, if the buffer capacity of the router 1 is 1 megabyte, the allocation result is as shown in Table 3.

  It should be noted that an arbitrary buffer capacity can be allocated to each user when there are limited users depending on the communication contract contents, device types, or the like, or when it is not necessary to be fair.

  According to the communication control method of the present invention, the bandwidth of each user's flow can be allocated fairly by the simple method shown in this paper without performing a complicated scheduling method.

In addition, it is possible to allocate the buffer capacity of the packet in consideration of only the user who is communicating, which is difficult with the scheduling method. For example, when a device accommodating 1000 users has a buffer capacity of 1 megabyte and there are only 10 users communicating at a certain time, the scheduling method uses a capacity of 1 kilobyte for each user. Buffer capacity allocation.
However, if the communication control method of the present invention is used, a buffer capacity of 100 kilobytes can be allocated.

    In addition, when the communication speed is improved or the buffer capacity is increased, the present invention can be implemented by modifying or adding a program.

  Further, in the above description, the case where a control program for controlling the router, which is a communication control device, is stored in advance in a ROM or the like has been described. However, control is performed on recording media such as various magnetic disks, optical disks, and memory cards. It is also possible to record the program for use in advance, read from these recording media, and install the program.

  In addition, a communication interface may be provided so that the control program can be downloaded, installed, and executed via a network such as the Internet or a LAN. By configuring in this way, it becomes possible to configure a software control function with higher functionality or a more reliable communication control device, and the possibility of industrial use is high.

FIG. 1 is a diagram showing a structure of a communication control device (router) provided with a policer. FIG. 2 is a diagram for explaining the structure of the policer. FIG. 3 is a diagram illustrating a configuration example of a flow table. FIG. 4 is a flowchart showing the operation of the policer policing processing unit. FIG. 5 is a diagram for schematically explaining the scheduling method. FIG. 6 is a diagram for schematically explaining a polishing method using a token bucket model. FIG. 7 is a diagram showing a network configuration using a router having a policer.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Router 2-1-2-n Input / output port 3 Switch fabric 4 Crossbar switch 5 Packet classification mechanism 6 Traffic adjustment mechanism 7 Packet adjustment mechanism 8 Policer 9 Discard processing part (dropper)
DESCRIPTION OF SYMBOLS 10 Connection arbitration circuit 11 Queue manager 12-1 to 12-n Queue 13 Scheduler 14 Input / output port 15 Traffic control mechanism 16 Buffer capacity 17 Out link 18 Policing processing unit 19 Flow table 20 Rate storage unit 21 Active counter unit 22 Burst storage unit 23 Flow State Storage Unit 24 Timer Unit 25-1 to 25-n Information Source 26 Token 27 Token Bucket 28 Packet 29 LAN
30-1 to 30-5 User 31, 34 Transmission path 32-1 to 32-n Flow 33 Network processor

Claims (10)

A communication control method that performs transmission by performing bandwidth control by policing processing on input packet data,
Classifying the input packet data into flows for at least one user;
A policing step for allocating all the bandwidths allocated for the transmission to the flows to be transmitted in parallel to the flows to be actually transmitted ,
The flow is assigned a weight according to the bandwidth ratio,
The policing step includes:
When the weight of the flow of the flow number i that is actually communicating is W _i , the total value of the weights of the entire flow that is currently communicating is A, and the packet data length is L,
The theoretical expected value (TAT value) of the arrival time of the next packet of the flow number i is calculated from the formula TAT _i = TAT _i + L × A / W _i
By performing the policing process using the predicted value,
A communication control method, comprising: allocating a bandwidth based on a ratio of W _i / A to a flow that is actually communicating . 2. The communication control method according to claim 1, wherein when at least one user is assigned to the flow to be actually transmitted, a transmission rate of data transmitted by the at least one user is varied. The communication control method according to claim 1 or 2 , wherein a packet capacity to be allocated is varied for each of the at least one user. A communication control device that performs transmission by performing bandwidth control by policing processing on input packet data,
A packet classification mechanism for classifying the input packet data into flows for at least one user;
A policing processing unit that allocates the entire bandwidth allocated for transmission to a plurality of flows that can be transmitted in parallel, to the flow to be actually transmitted ,
The flow is assigned a weight according to the bandwidth ratio,
The polishing processing unit
When the weight of the flow of the flow number i that is actually communicating is W _i , the total value of the weights of the entire flow that is currently communicating is A, and the packet data length is L,
The theoretical expected value (TAT value) of the arrival time of the next packet of the flow number i is calculated from the formula TAT _i = TAT _i + L × A / W _i
By performing the policing process using the predicted value,
A communication control apparatus characterized by allocating a bandwidth based on a ratio of W _i / A to a flow that is actually communicating . 5. The communication control apparatus according to claim 4 , wherein when allocating the flow to be actually transmitted for each at least one user, a transmission rate of data transmitted by the at least one user is varied. 6. The communication control apparatus according to claim 4 , wherein a packet capacity to be allocated is varied for each at least one user. A control program for controlling a communication control device that performs transmission by performing bandwidth control by policing processing on input packet data,
Classifying the input packet data into at least one user flow;
A policing process step of allocating the entire bandwidth allocated for the transmission to a plurality of flows that can be transmitted in parallel to the flow to be actually transmitted ;
The flow is assigned a weight according to the bandwidth ratio ,
The policing processing step includes:
When the weight of the flow of the flow number i that is actually communicating is W _i , the total value of the weights of the entire flow that is currently communicating is A, and the packet data length is L,
The theoretical expected value (TAT value) of the arrival time of the next packet of the flow number i is calculated from the formula TAT _i = TAT _i + L × A / W _i
By causing the policing processing step to be executed using the predicted value,
A control program that causes a flow that is actually communicating to perform bandwidth allocation based on a ratio of W _i / A. 8. The control program according to claim 7 , wherein when at least one user is assigned to the flow to be actually transmitted, a transmission rate of data transmitted by the at least one user is varied. 9. Wherein at least every one user, characterized by varying the packet capacity assigned, the control program according to claim 7 or 8. A computer-readable recording medium on which the control program according to any one of claims 7 to 9 is recorded.

Images