JP3649661B2 - Packet scheduling method and packet scheduling apparatus - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a packet scheduling method and a packet scheduling apparatus, and is used, for example, in various routers in an autonomous system (AS) provided in an IP (Internet Protocol) network to determine the processing order of input packets.
[0002]
[Prior art]
For example, various types of packets are sequentially input to a router provided in an IP network or the like. A plurality of queues (queues) are provided inside a router that processes various types of packets, and input packets are stored in queues corresponding to the types and then processed.
[0003]
Such a process for determining the order in which packets stored in a plurality of queues are processed is called packet scheduling. The following techniques are conventionally known as packet scheduling methods.
(1) Round Robin: Packets in the first queue, packets in the second queue, packets in the third queue,... The number of packets to be processed at one time is determined in advance (generally one). Processing is skipped for queues where there are no packets to be processed.
[0004]
(2) WRR (Weighted Round Robin): When the packet weight is different from the “weight” for each packet class (queue), use the “normalized weight” obtained by dividing the weight by the average packet length. The number of packets corresponding to is processed by the round robin.
(3) DRR (Deficit Round Robin): In the WRR scheme, it is devised not to require information on the average packet length in advance. The details are processed in the following procedure. A certain “reference amount” is given for each class (queue), and if the packet length at the head of a certain queue is equal to or smaller than the reference amount, the packet is processed. The reference amount is reduced by the packet length. In this way, the queue is processed continuously until the reference amount is less than the packet length to be processed next. When it is not satisfied, the processing target moves to the next queue in the round robin method. When the head packet length of all queues becomes larger than the reference amount of each queue, a reference amount is newly added to all the queues.
[0005]
(4) GPS (Generalized Processor Sharing): This is a scheduling method in which servers having an infinite processing capacity for N queues in which packets exist are simultaneously processed at a service rate of 1 / N. When there is no packet to be processed in a certain queue, the processing capacity for that is equally allocated to the other queues. This is an ideal method that cannot be realized in practice.
[0006]
(5) WFQ (Weighted Fair Queueing): This is a method for approximating the GPS system. In GPS, in parallel, infinitesimal processing is performed in parallel. The packet processing end scheduled time is calculated bit by bit, and processing is performed in order from the packet with the earlier time.
(6) CB-WFQ (Class-based Weighted Fair Queueing): A weight value is fixedly assigned to each queue.
[0007]
For example, Japanese Patent Laid-Open No. 9-83547 discloses that scheduling is performed based on the weight set for each packet queue and the packet accumulation amount of the packet queue.
Further, when selecting a packet scheduling method, the selection is generally made in consideration of the following matters.
[0008]
(A) What and how to realize by scheduling (for example, wanting to finish the processing within the upper limit of the difference in priority, fairness, and delay time).
(B) How to define priority and whether it is actually realized.
(C) Whether there is a need to consider fairness between classes, and how to consider it if necessary.
[0009]
(D) How easy is it to actually realize it in the device?
The most convenient method is selected in consideration of the combination of the above items.
[0010]
[Problems to be solved by the invention]
For example, in an IP network, various types of packets are mixed and transmitted in order to carry information of various types of applications. Application types include voice (VoIP), video (Video on Demand and video conference), Telnet, X-Window, WWW, electronic mail, FTP, and the like.
[0011]
Packets that carry voice and image application information are called "stream traffic". A packet carrying application information such as Telnet or X-Window is called “interactive traffic”. Further, a packet carrying application information such as WWW, electronic mail, and FTP is called “bulk transfer type traffic”.
[0012]
For “stream type traffic” and “interactive type traffic”, the real-time property is important, and if the delay time in the middle of transmission increases, the transmission quality deteriorates remarkably. For “bulk transfer traffic”, the real-time property is not so important.
When the CB-WFQ method is adopted, a large weight is allocated in advance to a queue that processes “stream traffic” and “interactive traffic”, and a small weight is allocated to a queue that processes “bulk transfer traffic”. By doing so, it is possible to ensure the quality for applications where real-time requirements are severe.
[0013]
However, when the weight of each queue is fixedly determined as in the CB-WFQ method, the quality of a specific application using a queue to which a small weight is assigned may be significantly reduced. That is, when a specific queue is used frequently and many packets arrive within a short time compared to the bandwidth allocated to the queue, the delay time in the queue is significantly increased.
[0014]
Further, when using the technique disclosed in Japanese Patent Laid-Open No. 9-83547, scheduling is dynamically performed according to the amount of packets stored in the packet queue, so that the priority of each class of packets changes sequentially. . Therefore, the bandwidth cannot be guaranteed for a specific class of packets having high real-time characteristics. In addition, the size of the bandwidth allocated to a specific class may become too small.
[0015]
In the packet scheduling method and the packet scheduling apparatus, the present invention effectively uses all the available bandwidths to prevent the deterioration of the quality of applications that are severe in real-time characteristics, and transmits even applications with small real-time requirements. The purpose is to prevent the delay time from increasing.
[0016]
[Means for Solving the Problems]
  Claim 1 classifies input packets into a plurality of classes, classifies the plurality of classes into priority groups in which real time required for processing is important and non-priority groups other than the priority groups, and sequentially A packet scheduling method for holding an input packet as a queue provided for each of the plurality of classes, and determining an order of packets to be taken out of the queue for processing, wherein the packet is assigned to the non-priority group When the allocated bandwidth is too small and there is an unprocessed non-priority class (hereinafter referred to as the first non-priority class), the queue for storing the packets of the unprocessed non-priority class First, assigning a bandwidth BWα obtained by first multiplying all bandwidths BWt by X (X is a coefficient);
  A second band BWr that is sufficient to satisfy the required delay time of each priority class in the priority group is obtained by subtracting the band BWα allocated to the first non-priority class from the total band BWt. For the second non-priority class obtained by removing the first non-priority class from the non-priority class belonging to the non-priority group.βTo bandwidth BWrAnd a third step of assigning a band BWz obtained by subtracting.
[0017]
  In claim 1,Band BWα is allocated to the first non-priority class. This facilitates the processing of queues where the allocated bandwidth is too small and processing has been delayed.
  In addition, the queue holding the packets of each priority class belonging to the priority group is allocated a bandwidth BWr sufficient to satisfy the request delay time of each priority class.It is guessed.
  Further, the band BWz is assigned to the second non-priority class.
  Therefore, a priority group whose real time required for processing is important and a non-priority group other than the priority group are efficiently processed.
[0018]
  A second aspect of the present invention provides the packet scheduling method according to the first aspect, wherein whether or not the first non-priority class exists is determined., SpecialPeriodBy identifying for each class belonging to the non-priority group whether or not there is a class in which the bandwidth allocated duringHolds packets of each class belonging to the non-priority groupeachQueuePriority for extracting the packet fromIs performed based on a weight according to a difference in usage frequency detected in the most recent specific period.
[0020]
  3. The packet scheduling method according to claim 1, wherein, in the first step, the usage frequency of each class of the non-priority group is determined by the number of packets arriving in the specific period and an average packet of arriving packets. Based on lengthDynamicallyIt is characterized by detecting.
[0021]
  ContractIn claim 3, since the frequency of use of each class is detected based on the number of packets arriving in the specific period and the average packet length of the arriving packets, each packet length of the arriving packet varies. Can also detect the usage frequency of each class.
[0022]
  According to a fourth aspect of the present invention, the input packets are classified into a plurality of classes, and the plurality of classes are classified into a priority group in which real time required for processing is important and a non-priority group other than the priority group, and sequentially A packet scheduling apparatus for storing packets input to the plurality of classes as a queue provided for each of the plurality of classes and determining an order of packets taken out from the queue for processing, the non-priority group When the size of the bandwidth allocated to is too small and there is an unprocessed non-priority class (hereinafter referred to as the first non-priority class), the queue for storing the packets of the unprocessed non-priority class is stored in the queue. On the other hand, first band allocating means for allocating a band BWα obtained by first multiplying all bands BWt by X (X is a coefficient), and from all bands BWt Second band allocating means for allocating a band BWr sufficient to satisfy the required delay time of each priority class in the priority group to the band BWβ obtained by subtracting the band BWα allocated to the first non-priority class For the second non-priority class obtained by removing the first non-priority class from the non-priority class belonging to the non-priority group, the bandwidth BWβTo bandwidth BWrAnd a third band allocating unit that allocates a band BWz obtained by subtracting.
[0023]
In the packet scheduling apparatus according to the fourth aspect, the packet scheduling can be performed similarly to the first aspect.
[0024]
DETAILED DESCRIPTION OF THE INVENTION
An embodiment of a packet scheduling method and a packet scheduling apparatus according to the present invention will be described with reference to FIGS. This form corresponds to all the claims.
[0025]
  FIG. 1 is a flowchart showing packet scheduling processing in this form. FIG. 2 is a block diagram showing a configuration of a router that implements the present invention. FIG. 3 is a time chart showing the timing of processing in this form.
  In this form,The first bandwidth allocation means, the second bandwidth allocation means, and the third bandwidth allocation means are:Traffic characteristic amount calculation unit 16 (S15, S19, S22, S25)And the processing class instruction unit 18.
[0026]
In this embodiment, it is assumed that the packet scheduling method and the packet scheduling apparatus of the present invention are applied to the router of FIG. 2 connected to the IP network.
Referring to FIG. 2, this router includes a packet information collection unit 10, a class determination unit 11, a header reading measurement unit 12, a packet header analysis / calculation unit 13, a routing table 14, a header write control unit 15, and a traffic characteristic amount calculation unit. 16, an allocated bandwidth calculation unit 17, a processing class instruction unit 18, a database (DB) 19, a scheduling unit 20, a clock 31 and a timer 32.
[0027]
The scheduling unit 20 includes N queue buffers 21 (1) to 21 (N) and a packet processing unit 22. The queue buffer 21 can be composed of, for example, a FIFO (first in first out) memory.
Packets arriving at this router from the outside are input to the packet information collecting unit 10, accumulated in any one queue buffer 21 by the scheduling unit 20, processed by the packet processing unit 22, and output to the outside.
[0028]
The packet information collection unit 10 extracts header information including the destination, class, and packet length from each packet and notifies the class determination unit 11 of the header information. The class determination unit 11 classifies the packet type into N or less classes based on the header information class. Also, the class of the priority group is distinguished from the class of the non-priority group.
The packet information collection unit 10 writes the input packet into one queue buffer 21 corresponding to the class classified by the class determination unit 11.
[0029]
The header information including the destination and class of each packet is input from the class determination unit 11 to the packet header analysis / calculation unit 13 through the header reading measurement unit 12.
The packet header analysis / calculation unit 13 refers to the routing table 14 based on the header information and acquires destination information. This destination information passes through the header write control unit 15 and is written in a packet on the queue buffer 21 as next router information.
[0030]
The traffic characteristic amount calculation unit 16 calculates the traffic characteristic amount for each specific section for each class of packet. The allocated bandwidth calculation unit 17 allocates the bandwidth of each class according to the characteristic amount obtained by the traffic characteristic amount calculation unit 16.
The bandwidth value allocated by the allocated bandwidth calculation unit 17 is input to the processing class instruction unit 18 as the weight of each class. Further, the processing class instruction unit 18 receives the number of processing packets for each specific section from the traffic characteristic amount calculation unit 16.
[0031]
The processing class instruction unit 18 instructs the packet processing unit 22 on the class of the packet to be processed by the packet processing unit 22. The packet processing unit 22 takes out the packet of the class instructed from the processing class instruction unit 18 from the head of the corresponding queue buffer 21 and processes it.
Details of the packet scheduling process for determining the order of packets taken out from the queue buffers 21 (1) to 21 (N) by the packet processing unit 22 of FIG. 2 will be described below.
[0032]
The packet class may be determined depending on the type of application, may be determined at the time of user contract, or may be determined at the start of communication.
Application types include voice (VoIP), video (Video on Demand and video conference), Telnet, X-Window, WWW, electronic mail, FTP, and the like.
[0033]
Packets that carry voice and image application information are called "stream traffic". A packet carrying application information such as Telnet or X-Window is called “interactive traffic”. Further, a packet carrying application information such as WWW, electronic mail, and FTP is called “bulk transfer type traffic”.
[0034]
For “stream type traffic” and “interactive type traffic”, the real-time property is important, and if the delay time in the middle of transmission increases, the transmission quality deteriorates remarkably. For “bulk transfer traffic”, the real-time property is not so important.
Therefore, in this packet scheduling process, the packet class is classified into a priority group and a non-priority group. Packet classes corresponding to “stream traffic” and “interactive traffic” are classified into priority groups because real-time characteristics are important. Further, the packet class corresponding to “bulk transfer type traffic” is classified into a non-priority group because the real-time property is not important. Each class belonging to the priority group is called a priority class, and each class belonging to the non-priority group is called a non-priority class.
[0035]
The packet scheduling process shown in FIG. 1 represents a bandwidth allocation procedure for queues (the contents of each queue buffer 21) that process packets of each class. The processing priority of the packets accumulated in each queue is determined according to the size of the bandwidth allocated to each queue.
Further, the packet scheduling process shown in FIG. 1 is executed at the timing shown in FIG. That is, when the current time is t, information is collected for the previous information collection section A1, calculation is performed in the calculation section A2, and bandwidth is allocated to the control target section A3.
[0036]
If the packet arrives at the input of the router in FIG. 2, the packet scheduling process proceeds from step S12 to step S13 in FIG. In step S13, the class of the arrived packet is identified. In the next step S14, the arrived packet is transferred to one queue buffer 21 associated with the class. In step S15, the number of packets arriving for each class and the average packet length are calculated.
[0037]
When the predetermined timing for allocating the bandwidth to each class is reached, the process proceeds from step S11 to S16. Steps S16 to S19 are processes for guaranteeing a minimum bandwidth for a non-priority class packet.
In step S16, it is identified whether or not there is a non-priority class in a certain monitoring section (between t-KT and t) where the allocated bandwidth is too small to process the packet at all. When it exists, it progresses to step S18, and when it does not exist, it progresses to step S17.
[0038]
In step S17, the value of the band BWα (bps) is set to zero.
In step S18, the result of multiplying the entire bandwidth BWt (bps) of the router by a coefficient X (for example, 0.1) is set to the value of the bandwidth BWα. The coefficients X and K are obtained by calculation so that the delay time of each class does not exceed a certain value.
In step S19, the bandwidth BWα is assigned to the non-priority class that could not process the packet.
[0039]
When there are a plurality of non-priority classes that could not process a packet, a part of the bandwidth BWα is assigned to each non-priority class so that the delay time of each class does not exceed a certain value.
In step S20, the result of subtracting the band BWα from the entire band BWt is determined as the value of the band BWβ.
[0040]
In step S21, for the priority class packet, the stay time Tisys (A1) in the node (router) of the packet observed in the information collection section A1 is calculated for each class (only the priority class) by the following equation.
Tisys (A1) = xi (A1) + ρi (A1) xi (A1) / (2 (1-ρi (A1))) (1)
However,
xi (A1) = 8 pi (A1) / BW (Ci, A1): Average processing time (sec)
pi (A1): Average packet length (bytes)
BW (Ci, A1): Band size allocated to the i-th class
ρi (A1) = λi (A1) xi (A1): Usage rate
λi (A1) = qi (A1) / nT: arrival rate
qi (A1): Number of packets of the i-th class that arrived in the information collection section A1
In step S22, bandwidth allocation is updated for each class for the priority class. Specifically, the size of the band BW (Ci, A3) is assigned so that the stay time Tisys (A1) of the packet obtained in step S21 is smaller than the delay time di requested by each class. That is, for the priority class, a band that satisfies the condition of the delay time di is given to the control target section A3 based on information obtained by collecting the last n periods (A1).
[0041]
  Accordingly, the bandwidth is allocated to all the priority classes so as to substantially satisfy the delay time di required by the priority class. In addition, for a priority class packet that cannot satisfy the delay time di, the call reception is controlled so as to reject the reception of such a call in advance.
  In step S23, first, a sum BWr of the bandwidth BW (Ci, A3) assigned to all classes belonging to the priority group in step S22 is obtained.
  In step S24,The result of subtracting the band BWr from the band BWβ is obtained as the remaining band BWz.
[0042]
  Step S25In the non-priority class, the remaining bandwidth BWz is assigned for each class. Specifically, each kth class is determined based on the number of arrival packets N (Ck, A1) and average packet length PL (Ck, A1) (non-priority class only) collected in the information collection section A1. The Ck band BW (Ck, A3) is obtained using the following equation.
BW (Ck, A3) = BWz · uk · vk / (sum of uk and vk of non-priority class) (2)
However,
uk = N (Ck, A1) (non-priority class only)
vk = PL (Ck, A1) (non-priority class only)
  That is, for the non-priority class packets, the bandwidth BWz is allocated to each class at a rate proportional to the amount of packet information (number of packets × average packet length) arriving in the information collection section A1. For this reason, more and more bandwidth is allocated to the non-priority class as the frequency of use increases. In addition, as the usage frequency decreases, the allocated bandwidth decreases.
[0043]
The size of the bandwidth BW (Cj) (j = 1 to N) allocated to each class by the processing of FIG. 1 corresponds to the “weight” at the time of router scheduling, and is directly linked to the priority in the control target section A3. .
That is, the priority weight W (Cj, A3) of the class Cj in the control target section A3 is given by the following equation.
[0044]
W (Cj, A3) = BW (Cj, A3) / BWt (3)
However, for each class of the non-priority group, the queue is selected according to the above weight and the packet is processed only when there is no packet in the queue of each class of the priority group.
Next, a specific example of bandwidth allocation will be described.
[0045]
(Specific example 1)
In this example, the following state is assumed.
BWt: 12Mbps
BWα: 0
T: 5 seconds
n: 1
m: 1
Number of classes N: 3
Class 1: Priority class
Class 2, Class 3: Non-priority class
Current time: 0
Further, the following state is assumed as the usage state of each class in the information collection section A1 (−5, 0).
[0046]
Number of class 1 arrival packets: 2
Number of class 2 arrival packets: 2
Number of class 3 arrival packets: 6
Class 1 average packet length: 100
Class 2 average packet length: 200
Class 3 average packet length: 300
Here, assuming that the bandwidth BWr is 8 Mbps, the bandwidth assigned to each class for the control target section A3 (5, 10) is as follows.
[0047]
Class 1: 8 Mbps
Class 2: (12-8) x 2 x 200 / (2 x 200 + 6 x 300) = 0.727 Mbps
Class 3: (12-8) x 6 x 300 / (2 x 200 + 6 x 300) = 3.273 Mbps
(Specific example 2)
In this example, in a section with an integer multiple (KT) of the observation period T, the number of arrivals of packets in the priority class is too large, and the allocation of bandwidth to the non-priority group is reduced, and a certain class belonging to the non-priority group (not yet) Assume that no processing class packet is processed.
[0048]
In this case, for each section (KT), the X (%) band of the entire band in the next control target section A3 is assigned to the unprocessed class for processing. The coefficients K and X are obtained by calculation so that the delay time of each class does not exceed a certain value.
That is, when an unprocessed non-priority class is detected, the bandwidth BWα is assigned to an unprocessed non-priority class in step S19 in FIG.
[0049]
Here, the following state is assumed.
BWt: 12Mbps
T: 5 seconds
n: 1
m: 1
Number of classes N: 3
Class 1: Priority class
Class 2: Non-priority class
Class 3: Non-priority class (unprocessed class)
K: 2
X: 10 (%)
Current time: 0
Further, the following state is assumed as the usage state of each class in the information collection section A1 (−5, 0).
[0050]
Number of class 1 arrival packets: 9
Number of class 2 arrival packets: 1
Number of class 3 arrival packets: 0
Class 1 average packet length: 100
Class 2 average packet length: 200
Class 3 average packet length: 300
Here, it is assumed that the bandwidth BWr is 11 Mbps and no class 3 packet is processed in the section (−10, 0). In this case, the bandwidth allocated to each class for the control target section A3 (5, 10) is as follows.
[0051]
Class 1: 12-1.2 = 10.8Mbps
Class 2: 0 Mbps
Class 3: 12 × 10/100 = 1.2 Mbps
That is, a bandwidth of (1.2 Mbps) is preferentially assigned to class 3, which is an unprocessed class, so that the packet can be processed in the next control section.
[0052]
Therefore, in this form, a sufficient bandwidth can be secured in advance for the class having high real-time characteristics to satisfy the required delay time. In addition, for classes with a large amount of usage (determined by the number of packets) in the most recent time interval, priority control for increasing throughput and shortening delay time is dynamically performed within a very short time close to real time. Can be realized.
[0053]
Also, a minimum bandwidth can be allocated to a class with a small number of passing packets that is not frequently used so that the delay time in each node (router) does not exceed a certain limit value.
[0054]
【The invention's effect】
According to the present invention, it is possible to allocate a bandwidth so that a packet of a frequently used class is prioritized. In addition, the bandwidth of the packet of the class can be allocated so as to satisfy the required quality of the application with high real-time property. Further, even for a class having low real-time characteristics, the average throughput for a relatively long time can be set to a certain value or more.
[0055]
By adopting the apparatus of the present invention, it becomes possible to allocate a bandwidth according to the frequency of use to a user who uses an IP network or an application to be used. Can be activated.
[Brief description of the drawings]
FIG. 1 is a flowchart illustrating packet scheduling processing according to an embodiment.
FIG. 2 is a block diagram showing a configuration of a router for implementing the present invention.
FIG. 3 is a time chart illustrating processing timings according to the embodiment.
[Explanation of symbols]
10 Packet information collection unit
11 Class decision part
12 Header reading measurement unit
13 Packet header analysis / calculation section
14 Routing table
15 Header write controller
16 Traffic characteristic calculation section
17 Allocated bandwidth calculator
18 Processing class indicator
19 Database
20 Scheduling Department
21 Queue buffer
22 Packet processing part
31 clock
32 timer

Claims (4)

The input packets are classified into a plurality of classes, and the plurality of classes are classified into a priority group in which real time required for processing is important and a non-priority group other than the priority group. A packet scheduling method for determining a sequence of packets held as a queue provided for each of the plurality of classes and taken out of the queue for processing,
If the bandwidth allocated to the non-priority group is too small and there is an unprocessed non-priority class (hereinafter referred to as the first non-priority class), the packets of the unprocessed non-priority class are stored. A first step of assigning a bandwidth BWα obtained by first multiplying all bandwidths BWt by X (X is a coefficient) to the queue to be
A second bandwidth BWr that is sufficient to satisfy the required delay time of each priority class in the priority group is obtained by subtracting the bandwidth BWα allocated to the first non-priority class from the total bandwidth BWt. And the steps
A third non-priority class obtained by subtracting the first non-priority class from the non-priority class belonging to the non-priority group is assigned a band BWz obtained by subtracting the band BW r from the band BW β . A packet scheduling method comprising the steps of: The packet scheduling method according to claim 1, wherein
Said first whether the non-preferential class exists judgment, too small bandwidth allocated between between Japanese periodical, whether the non-priority group or class that could not be processed at all the packet exists By identifying each class belonging to
The priority for extracting the packet from each queue that holds packets of each class belonging to the non-priority group is determined based on the weight according to the difference in use frequency detected in the most recent specific period. A characteristic packet scheduling method. The packet scheduling method according to claim 1, wherein
In the first step,
A packet scheduling method, wherein the usage frequency of each class of the non-priority group is dynamically detected based on the number of packets arriving in the specific period and the average packet length of the arriving packets. Packets that are input sequentially by classifying the input packets into a plurality of classes, and classifying the plurality of classes into priority groups in which real time required for processing is important and non-priority groups other than the priority groups A packet scheduling device for determining the order of packets taken out of the queue for processing,
If the bandwidth allocated to the non-priority group is too small and there is an unprocessed non-priority class (hereinafter referred to as the first non-priority class), the packets of the unprocessed non-priority class are stored. First bandwidth allocating means for allocating a bandwidth BWα obtained by first multiplying the entire bandwidth BWt by X (X is a coefficient) to the queue to be
A second bandwidth BWr that is sufficient to satisfy the required delay time of each priority class in the priority group is obtained by subtracting the bandwidth BWα allocated to the first non-priority class from the total bandwidth BWt. Bandwidth allocation means,
A third non-priority class obtained by subtracting the first non-priority class from the non-priority class belonging to the non-priority group is assigned a band BWz obtained by subtracting the band BW r from the band BW β . Bandwidth allocation means;
A packet scheduling apparatus comprising:

Images