CN1254053C - Weighing polling method realiznig minimum jitter of cell flux - Google Patents
Weighing polling method realiznig minimum jitter of cell flux Download PDFInfo
- Publication number
- CN1254053C CN1254053C CN 02155335 CN02155335A CN1254053C CN 1254053 C CN1254053 C CN 1254053C CN 02155335 CN02155335 CN 02155335 CN 02155335 A CN02155335 A CN 02155335A CN 1254053 C CN1254053 C CN 1254053C
- Authority
- CN
- China
- Prior art keywords
- formation
- cbr
- cell
- look
- numbering
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
Images
Landscapes
- Data Exchanges In Wide-Area Networks (AREA)
Abstract
The present invention relates to a weighted polling method realizing the minimum jitter of cell flow. In the method, each cell queue is first numbered; a search table for the uniform distribution of the numbers of the queues is then configured, and the cell of each queue is sent according to the sequence of the search table; consequently, the flow dispatching of the queues is realized. The present invention has the advantages that the present invention can be easily realized, the defects existing in a traditional weighted polling algorithm are solved, mask registers of the queues are not compared with each other during realization so that difficulty and complexity for realizing a weighted polling algorithm can be greatly decreased, the scale and the difficulty of the realization of logic devices are greatly decreased when the present invention is realized through using ASIC or FPGA, and flow jitter in short time, which is caused by the traditional weighted polling algorithm, is greatly reduced so that the jitter in short time is reduced to the lowest degree. In addition, only the table is searched in the present invention; thereby, a CPU is used for calculation at a high speed which is much higher than the speed of the existing weighted polling algorithm.
Description
Technical field
The present invention relates to a kind of WRR method that makes the cell flow realize minimum jitter, exactly, relate to a kind of when using traditional Weighted Round Robin, the shake of cell on the short time is excessive, and when using application-specific integrated circuit ASIC or on-site programmable gate array FPGA device to realize, the WRR method that the problem that the logic scale is too big is better solved belongs to the cell flow scheduling controlling technical field of asynchronous transfer mode and other professional constant bit rate CBR business.
Background technology
WRR (WRR) algorithm is mainly used on the flow scheduling of ATM and other professional CBR business.Now, the traditional Weighted Round Robin for control cell flow generally adopts following implementation method:
In order to calculate and to compare markers, the markers counter that every queue assignment is 16, initial value is 0x0000, being designated as 16 here when determining is in order to reduce the quantity of register, to accelerate relatively speed; The transmission time interval parameter of another key parameter-16 then is stored in the register parameters table, this parameter is to calculate according to the reserved bandwidth of formation and weights, the weights of formation and cell time spacing parameter need CPU to be configured, when carrying out configuration, write register.Adopt 16 time interval parameter, resolution is: 622M/64K=9.8K; The computing formula of weights is: WEIGHT=|R/9.8K|, R is the reservation speed of every formation in the formula.
Choose formation by poll at every turn, after finding a formation, does the value of checking this formation markers counter earlier meet the transmission condition: formation [n] Counter<formation [m] Counter? if it is eligible, just send a cell of this formation, and the counter of this formation added the time interval value of this formation, poll carries out computing next time again.Otherwise, just send the cell of another formation, and make the counter of its formation add the time interval value of this formation.If one the count value of formation arrives maximum 0xFFFF, just wait for the calculating of another formation.If all counters all arrive or during near maximum 0xFFFF, simultaneously all counter O resets, calculate again.
Referring to Fig. 1, this Figure illustrates that two formation CBR_A and CBR_B use respectively that weights are 7 and 4, the cell interval parameter is 4 and 7 to carry out the simplification scheduling signal of above-mentioned traditional Weighted Round Robin.In this traditional Weighted Round Robin, need the relatively size of these two marker registers for two formations, need also to judge whether these two registers approach 0xFFFF, so that being carried out zero clearing again, they handle.For a plurality of formations, the minimum value that just need find out each formation marker register judges which formation satisfies the transmission condition.
This traditional Weighted Round Robin can guarantee that the bandwidth of each formation in the long cycle is uniform.But when the bandwidth that a plurality of formations and each formation are arranged differed big, the resulting result of traditional Weighted Round Robin was bigger in the shake of going up in short-term.For the situation that two formations are only arranged, each formation shake of flow in short-term seems also not too obvious.If number of queues is more, shake is just very obvious in short-term.
Exemplify the situation that 16 formations are arranged of a simplification below: suppose that total bandwidth is 48M, always have 16 formation business (the priority height that the formation sequence number is little), the minimum bandwidth that needs in each formation is 1M.16 needed allocated bandwidth of formation of this simplification are as shown in the table:
The professional numbering of formation | The bandwidth of configuration | The professional numbering of formation | The bandwidth of configuration |
CBR_0 | 21M | CBR_8 | 1M |
CBR_1 | 13M | CBR_9 | 1M |
CBR_2 | 1M | CBR_10 | 1M |
CBR_3 | 1M | CBR_11 | 1M |
CBR_4 | 1M | CBR_12 | 1M |
CBR_5 | 1M | CBR_13 | 1M |
CBR_6 | 1M | CBR_14 | 1M |
CBR_7 | 1M | CBR_15 | 1M |
If adopt traditional weighting algorithm to carry out queue scheduling, the order of 48 scheduling that obtain is as shown in the table:
Sequence number | The formation numbering | Sequence number | The formation numbering | Sequence number | The formation numbering | Sequence number | The formation numbering |
1 | CBR_0 | 13 | CBR_12 | 25 | CBR_1 | 37 | CBR_0 |
2 | | 14 | CBR_12 | 26 | CBR_0 | 38 | CBR_1 |
3 | CBR_2 | 15 | CBR_14 | 27 | CBR_0 | 39 | CBR_0 |
4 | CBR_3 | 16 | CBR_15 | 28 | CBR_1 | 40 | CBR_0 |
5 | CBR_4 | 17 | CBR_0 | 29 | CBR_0 | 41 | CBR_1 |
6 | CBR_5 | 18 | CBR_1 | 30 | CBR_0 | 42 | CBR_0 |
7 | CBR_6 | 19 | CBR_0 | 31 | CBR_1 | 43 | CBR_0 |
8 | CBR_7 | 20 | CBR_0 | 32 | CBR_0 | 44 | CBR_1 |
9 | CBR_8 | 21 | CBR_1 | 33 | CBR_1 | 45 | CBR_0 |
Sequence number | The formation numbering | Sequence number | The formation numbering | Sequence number | The formation numbering | Sequence number | The formation numbering |
10 | CBR_9 | 22 | CBR_0 | 34 | CBR_0 | 46 | CBR_1 |
11 | CBR_10 | 23 | CBR_1 | 35 | CBR_0 | 47 | CBR_0 |
12 | CBR_11 | 24 | CBR_0 | 36 | CBR_1 | 48 | CBR_0 |
Referring to the formation sending time slots that wherein is numbered CBR_0 and CBR_1 shown in Figure 2 one section local distribution situation on time shaft, can clearly be seen that from Fig. 2: the formation speed that is numbered CBR_0 and CBR_1 is quite big in the shake of going up in short-term.
Therefore, the shortcoming of traditional Weighted Round Robin mainly is that the shake in short-term of flow is bigger, and is just very unsuitable to shaking the application of having relatively high expectations in short-term for some are arranged.When using this conventional art to realize simultaneously, need be by repeatedly more just obtaining a minimum value result.When using logical device to realize, the resource that needs is very big; If use CPU to calculate, then speed is very slow.
Summary of the invention
The purpose of this invention is to provide a kind of WRR method that makes the cell flow realize minimum jitter, excessive to solve the shake in short-term that prior art exists, and when using ASIC or FPGA device to realize the too big problem of logic scale.
The object of the present invention is achieved like this: a kind of WRR method that makes the cell flow realize minimum jitter, and it is characterized in that: this method comprises following three steps:
A, each cell-queue is numbered;
B, the numbering of each cell-queue evenly distributed is configured to a look-up table;
C, send the cell of each formation, realize the flow scheduling of formation according to the cell-queue's number order that disposes in the look-up table.
In the described steps A, before each cell-queue is numbered, further comprise the following steps:
A1, definite minimum discharge that needs each formation of realization;
A2, determine operational total bandwidth on the transmission channel;
A3, according to operational total bandwidth on above-mentioned minimum discharge and the transmission channel, obtain the length L of needed look-up table: L=total bandwidth/minimum discharge.
In the described steps A 3, needed look-up table is the RAM memory of each cell-queue's numbering of the described configuration of storage.
The length of described look-up table is the storage depth of RAM memory, i.e. the memory space of memory.
In the described steps A 3, the length L of the needed look-up table that obtains according to computing formula L=total bandwidth/minimum discharge can a round numbers part, so that the length of this look-up table is the shortest; Also can be accurate to the some positions behind the decimal point to the result of calculation of the length L of look-up table, promptly strengthen the length of look-up table, make the flow of each formation obtain more accurate control.
Among the described step B, disposing an equally distributed look-up table of the numbering with each formation is the method that adopts pseudo random interleaving.
Among the described step B, in look-up table, evenly distribute when disposing the numbering of each formation, its configuration result should satisfy following three conditions: the number of times that each formation numbering occurs is directly proportional with the bandwidth of its configuration, the numbering of representing each formation is evenly distributed on the RAM storage space, if and with the look-up table expansion that generates, on any one pitch in whole look-up table, the number of times that each formation numbering occurs and the error of its number of relevant position in look-up table are no more than 1.
Among the described step C, send the cell of each formation, further comprise the following steps: according to the order of look-up table
C1, according to the look-up table of above-mentioned configuration order, determine the formation numbering that sends, send a cell of this formation simultaneously;
C2, each send a cell after, the address of look-up table is added 1, read the data of this new address in the look-up table again, determine that promptly the next formation that sends numbers, then send a cell of this formation;
The operation of above-mentioned steps C2 is carried out in C3, circulation, finishes until whole transmission of the cell of all formations.
For synchronous digital hierarchy SDH/ Synchronous Optical Network SONET carrying bag or based on the elongated message on the dress POS/ packet interface SPI of system sealed of SDH/SONET, be cut to the cell of fixed length earlier, carry out described each step again, can realize the flow scheduling of elongated message.
The present invention is a kind of WRR method that makes the cell flow realize minimum jitter, and the key problem in technology of this method is to adopt the equally distributed look-up table of the numbering with each formation, realizes the flow scheduling of formation.Advantage of the present invention is to realize simply, solved the defective of traditional Weighted Round Robin, when realizing, the marker register that need not compare each formation, can reduce to realize the difficulty and the complexity of Weighted Round Robin greatly, using ASIC or FPGA to realize greatly having reduced scale and difficulty that logical device is realized the time of the present invention; Greatly reduced simultaneously the shake of the flow in short-term that traditional Weighted Round Robin causes, making in short-term, shake is reduced to minimum level.In addition, because the present invention only tables look-up, if use CPU to calculate, its speed is very fast, is higher than the speed of existing Weighted Round Robin far away.The present invention can also be at SDH/SONET carrying bag or the bag (POS/SPI of system, Packet Over SDH/SONET/System PacketInterface) flow scheduling of the elongated message of realization on the interface, promptly earlier elongated message is cut into the cell of fixed length, use the inventive method then and dispatch.
Description of drawings
Fig. 1 is the simplification scheduling schematic diagram of traditional Weighted Round Robin.
Fig. 2 is that the formation speed that is numbered CBR_0 and CBR_1 among the embodiment with traditional Weighted Round Robin shown in Figure 1 is shaken very big situation schematic diagram in short-term.
Fig. 3 is that the present invention makes the cell flow realize the flow chart of the WRR method of minimum jitter.
Fig. 4 adopts two formations that are numbered CBR_0 and CBR_1 in the embodiments of the invention to shake the situation schematic diagram that obviously reduces in the speed that is going up in short-term.
Embodiment
The present invention is a kind of WRR method that makes the cell flow realize minimum jitter, this method is earlier each cell-queue to be numbered, dispose an equally distributed look-up table of the numbering with each formation again, and send the cell of each formation, thereby realize the flow scheduling of formation according to the order of look-up table.
Referring to Fig. 3, this method comprises following concrete steps:
(1) minimum discharge of definite each formation that need realize or the minimum step (Bwsmin) of flow control; 64K for example.
(2) determine operational total bandwidth (Bwtmax) on the transmission channel; The bandwidth of a transmission channel of for example bandwidth of an optical fiber, or system assignment: 622M.
(3) according to operational total bandwidth on above-mentioned minimum discharge and the transmission channel, obtain the length L of needed look-up table: L=total bandwidth (Bwtmax)/minimum discharge (Bwsmin), just the degree of depth of RAM memory (being address space) is 622M/64K=9800, in order to reduce ram space, can be to its L round numbers part as a result.
(4) each constant bit rate CBR business is numbered one by one,, then formation can be numbered 0~4095 if 4096 formations are arranged; Adopt method evenly each formation of distribution configuration numbering in operational total bandwidth space of pseudo random interleaving again, be about to the even configuration that distributes in 9800 address spaces of each formation numbering 0~4095, and these configuration datas are deposited in the look-up table; And the number of times that configuration result should meet following condition: A, each formation numbering appearance is directly proportional with the bandwidth of its configuration.As to dispose professional ground of the 0th CBR bandwidth be 128K, and the number of times that occurs in look-up table (being RAM) of formation 0 should be 2 times so.B, represent that the numbering of each formation should uniform distribution on ram space.On C, any one pitch in whole look-up table 9800 addresses, the number of times that each formation numbering occurs is consistent with its number of relevant position in look-up table, and its error is no more than 1.
(5) according to the look-up table order of above-mentioned configuration, determine the formation numbering of transmission, send a cell of this formation simultaneously;
(6) each send a cell after, the address of look-up table is added 1, read the data of this new address in the look-up table again, determine that promptly the next formation that sends numbers, then send a cell of this formation;
(7) operation of above-mentioned steps (6) is carried out in circulation, finishes until whole transmission of the cell of all formations.
Adopt the same embodiment in the background technology further to specify it below.Suppose that total bandwidth is 48M, always have 16 formation business (the priority height that the formation sequence number is little), the minimum bandwidth that needs in each formation is 1M.16 needed allocated bandwidth of formation of this simplification are as shown in the table:
The professional numbering of formation | The bandwidth of configuration | The professional numbering of formation | The bandwidth of configuration |
CBR_0 | 21M | CBR_8 | 1M |
CBR_1 | 13M | CBR_9 | 1M |
CBR_2 | 1M | CBR_10 | 1M |
CBR_3 | 1M | CBR_11 | 1M |
CBR_4 | 1M | CBR_12 | 1M |
CBR_5 | 1M | CBR_13 | 1M |
CBR_6 | 1M | CBR_14 | 1M |
The professional numbering of formation | The bandwidth of configuration | The professional numbering of formation | The bandwidth of configuration |
CBR_7 | 1M | CBR_15 | 1M |
Adopt (1) in the step of the present invention~way of (5) to obtain being used for the look-up table of queue scheduling, it is one 4 * 48 RAM, and the data in the RAM see the following form:
Table address | The formation numbering | Table address | The formation numbering | Table address | The formation numbering | Table address | The formation numbering |
0 | | 12 | | 24 | CBR_1 | 36 | |
1 | CBR_14 | 13 | CBR_0 | 25 | CBR_0 | 37 | CBR_4 |
2 | CBR_0 | 14 | CBR_1 | 26 | CBR_7 | 38 | CBR_0 |
3 | CBR_1 | 15 | CBR_0 | 27 | CBR_0 | 39 | CBR_1 |
4 | CBR_0 | 16 | CBR_10 | 28 | CBR_1 | 40 | CBR_3 |
5 | CBR_13 | 17 | CBR_1 | 29 | CBR_0 | 41 | CBR_0 |
6 | CBR_0 | 18 | CBR_0 | 30 | CBR_6 | 42 | CBR_1 |
7 | CBR_1 | 19 | CBR_9 | 31 | CBR_0 | 43 | CBR_0 |
8 | CBR_12 | 20 | CBR_0 | 32 | CBR_1 | 44 | CBR_2 |
9 | CBR_0 | 21 | CBR_1 | 33 | CBR_5 | 45 | CBR_0 |
10 | CBR_1 | 22 | CBR_0 | 34 | CBR_0 | 46 | CBR_1 |
11 | CBR_0 | 23 | CBR_8 | 35 | CBR_1 | 47 | CBR_0 |
This look-up table is consistent with the order of queue scheduling, that is to say, should be according to the order transmit queue of this look-up table.
According to above-mentioned look-up table, then as shown in Figure 4 for the distribution of formation sending time slots on time shaft that is numbered CBR_0 and CBR_1.Can see obviously that from Fig. 4 behind employing the present invention, CBR_0 and CBRCBR_1 formation significantly reduce in the speed shake of going up in short-term, the benefit that the cell of each formation evenly sends is conspicuous.The present invention simultaneously is a target program when adopting the way of tabling look-up to omit relatively, has significantly reduced the difficulty and the scale that realize with ASIC and FPGA device.
Method of the present invention is tested enforcement on the equipment of applicant's development, the result of enforcement is successful, has realized goal of the invention.
Claims (8)
1, a kind of WRR method that makes the cell flow realize minimum jitter, it is characterized in that: this method comprises following three steps:
A, each cell-queue is numbered;
B, the numbering of each cell-queue evenly distributed is configured to a look-up table;
C, send the cell of each formation, realize the flow scheduling of formation according to the cell-queue's number order that disposes in the look-up table.
2, WRR method according to claim 1 is characterized in that: in the described steps A, before each cell-queue is numbered, further comprise the following steps:
A1, definite minimum discharge that needs each formation of realization;
A2, determine operational total bandwidth on the transmission channel;
A3, according to operational total bandwidth on above-mentioned minimum discharge and the transmission channel, obtain the length L of needed look-up table: L=total bandwidth/minimum discharge.
3, WRR method according to claim 1 and 2 is characterized in that: in the described steps A 3, needed look-up table is the RAM memory of each cell-queue's numbering of the described configuration of storage.
4, WRR method according to claim 2 is characterized in that: the length of described look-up table is the storage depth of RAM memory, i.e. the memory space of memory.
5, WRR method according to claim 2, it is characterized in that: in the described steps A 3, the length L of the needed look-up table that obtains according to computing formula L=total bandwidth/minimum discharge can a round numbers part, so that the length of this look-up table is the shortest; Also can be accurate to the some positions behind the decimal point to the result of calculation of the length L of look-up table, promptly strengthen the length of look-up table, make the flow of each formation obtain more accurate control.
6, WRR method according to claim 1 is characterized in that: among the described step B, disposing an equally distributed look-up table of the numbering with each formation is the method that adopts pseudo random interleaving.
7, WRR method according to claim 1 is characterized in that: among the described step C, send the cell of each formation according to the order of look-up table, further comprise the following steps:
C1, according to the look-up table of above-mentioned configuration order, determine the formation numbering that sends, send a cell of this formation simultaneously;
C2, each send a cell after, the address of look-up table is added 1, read the data of this new address in the look-up table again, determine that promptly the next formation that sends numbers, then send a cell of this formation;
The operation of above-mentioned steps C2 is carried out in C3, circulation, finishes until whole transmission of the cell of all formations.
8, WRR method according to claim 1 and 2, it is characterized in that: for synchronous digital hierarchy SDH/ Synchronous Optical Network SONET carrying bag or based on the elongated message on the dress POS/ packet interface SPI of system sealed of SDH/SONET, be cut to the cell of fixed length earlier, carry out described each step again, can realize the flow scheduling of elongated message.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN 02155335 CN1254053C (en) | 2002-12-10 | 2002-12-10 | Weighing polling method realiznig minimum jitter of cell flux |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN 02155335 CN1254053C (en) | 2002-12-10 | 2002-12-10 | Weighing polling method realiznig minimum jitter of cell flux |
Publications (2)
Publication Number | Publication Date |
---|---|
CN1507212A CN1507212A (en) | 2004-06-23 |
CN1254053C true CN1254053C (en) | 2006-04-26 |
Family
ID=34235854
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN 02155335 Expired - Fee Related CN1254053C (en) | 2002-12-10 | 2002-12-10 | Weighing polling method realiznig minimum jitter of cell flux |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN1254053C (en) |
Families Citing this family (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN100359863C (en) * | 2004-12-31 | 2008-01-02 | 中兴通讯股份有限公司 | Method and apparatus for realizing weighted round-robin scheduling using pseudo-random sequence |
WO2009085276A1 (en) * | 2007-12-28 | 2009-07-09 | Bytemobile, Inc. | Tcp packet spacing |
CN102684983B (en) * | 2011-03-15 | 2016-08-03 | 中兴通讯股份有限公司 | A kind of cell scheduling method and apparatus |
CN103338159B (en) | 2013-06-19 | 2016-08-10 | 华为技术有限公司 | Polling dispatching implementation method and device |
CN110995598B (en) * | 2019-11-12 | 2022-03-04 | 芯创智(北京)微电子有限公司 | Variable-length message data processing method and scheduling device |
-
2002
- 2002-12-10 CN CN 02155335 patent/CN1254053C/en not_active Expired - Fee Related
Also Published As
Publication number | Publication date |
---|---|
CN1507212A (en) | 2004-06-23 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US7751404B2 (en) | Method, system, and computer program product for high performance bonding resequencing | |
JP5671150B2 (en) | Lockless buffer management scheme for telecommunications network applications | |
US6356951B1 (en) | System for parsing a packet for conformity with a predetermined protocol using mask and comparison values included in a parsing instruction | |
US6661774B1 (en) | System and method for traffic shaping packet-based signals | |
CN1881937A (en) | Method and device for making storage space dynamic distribution for multi queue | |
CN1097913C (en) | ATM throttling | |
US20040128463A1 (en) | Apparatus and method for controlling memory allocation for variable size packets | |
US20050268037A1 (en) | Cache hit ratio estimating apparatus, cache hit ratio estimating method, program, and recording medium | |
CN103560980B (en) | For the packet of internal/external storage device and the method and apparatus of byte count | |
US20080126451A1 (en) | Allocating space on data storage devices in proportion to weights associated with the devices | |
US8312243B2 (en) | Memory management in network processors | |
CN1643499A (en) | Thread signaling in multi-threaded network processor | |
JP2006500830A5 (en) | ||
CN1254053C (en) | Weighing polling method realiznig minimum jitter of cell flux | |
CN1925455A (en) | Message speed limiting method based on token barrel | |
CN1577253A (en) | EDF scheduling method | |
CN1565143A (en) | Distributed transmission of traffic flows in communication networks | |
CN1359067A (en) | Method for making data in circular buffer effective | |
CN111881165B (en) | Data aggregation method and device and computer readable storage medium | |
CN1669276A (en) | Method for and system for controlling bandwidth allocation | |
US20060245443A1 (en) | Systems and methods for rate-limited weighted best effort scheduling | |
CN100344103C (en) | Method for uniform distribution of physical layer data transmitting time slot in whole time domain | |
CN102377641A (en) | Realization method for token bucket algorithm | |
CN113721858B (en) | Quick reading and writing method and device for memory, electronic equipment and storage medium | |
CN1841978A (en) | Method and apparatus for realizing multipath signal re-timing |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
C17 | Cessation of patent right | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20060426 Termination date: 20111210 |