CN100490383C - A high-speed Crossbar scheduling method for supporting multipriority - Google Patents
A high-speed Crossbar scheduling method for supporting multipriority Download PDFInfo
- Publication number
- CN100490383C CN100490383C CNB031583903A CN03158390A CN100490383C CN 100490383 C CN100490383 C CN 100490383C CN B031583903 A CNB031583903 A CN B031583903A CN 03158390 A CN03158390 A CN 03158390A CN 100490383 C CN100490383 C CN 100490383C
- Authority
- CN
- China
- Prior art keywords
- request
- input
- output
- crossbar
- iteration
- 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
This invention discloses a multi-priority high-speed crossbar dispatching strategy, which comprises queuing strategy and crossbar dispatching algorism and comprises the following steps: to store the group only in input end and divide the arrived group into signal units with certain length and to put them into queue with different priority; each input or output end has one cycle priority arbitration device; the dispatching algorism is co-executed by the devices with two phases of require and permit; finally to align crossbar and dispatching output signal element according to connection information.
Description
Technical field the present invention relates to a kind of high speed crossbar dispatching method of supporting multipriority, mainly is applicable to high speed, multiport, jumbo ip router/switch, and can provide service quality (QoS) to guarantee.
Background technology ip router/switch is a network interconnection device important in the internet, growth at full speed along with Intemet scale and capacity, and multiple services development, require router not only to support multiport, high-speed link speed, possess jumbo disposal ability, also want the QoS that can provide certain to guarantee, be in particular in the different demands of aspects such as satisfying throughput, bandwidth, time delay.In router, switching network is used to set up the data path between input and the output, is responsible for forwarding of data, is the key factor of restriction router port speed and capacity.
In traditional, the low speed router adopts the switching network (comprising shared buffer memory) of output work queue mostly, though they have good QoS performance, but the speed that requires switching network is N (N refers to the input port number) times of link rate, and in the internet main line link rate often very high (as OC-192,10GE), switching network is difficult to reach the speed of tens of Gbps, cause the system expandability poor, can't satisfy the growing demand of Internet.
Crossbar based on input queue is a kind of switching network of fixed length fast, the speed of only requiring switching network is identical with link rate, and crossbar has and is simple and easy to realize, clog-free etc. be widely used in advantage in the design of high speed router/switch
[1] [2] [3]In this switch network architecture, grouping is only stored at input, export by crossbar through scheduling, therefore, input queuing policy and crossbar dispatching algorithm are the key factors that influences performances such as throughput of system, time delay, two problems that must solve when also being high speed router/switch design.For the former, through former study, the problem that input rank descends for fear of the switching network throughput of bringing owing to HOL obstruction (head of line blocking)
[4], generally adopt virtual output work queue technology (VOQ:virtual output queueing) to eliminate HOL and block, promptly each input is safeguarded an independent FIFO (first in first out) formation for each output
[5]In order to support multipriority, each input is that each priority of each output is safeguarded an independent fifo queue.We are referred to as the VOQ formation by system.
For another problem---the crossbar dispatching algorithm, be described to the bigraph (bipartite graph) matching problem in the graph theory usually, promptly solve the competition of I/O end, avoid the transmission of cell and receive conflict.In order to realize the convenience in exchange at a high speed and the control, the data cell that crossbar handles is the cell of regular length, and a cell is got 64 byte lengths usually, and the time interval of a cell of crossbar exchange is called a time slot.Each time slot of dispatching algorithm is carried out once, the closure in execution result control crossbar crosspoint, and the foundation input is connected with output, and the cell in the input rank is sent to corresponding output by these connections.The multipriority scheduling is that a kind of simple and effective QoS guarantees mechanism.Support the crossbar dispatching algorithm of multipriority must satisfy following three conditions simultaneously: (a) each input sends a cell at the most in a time slot; (b) each output receives a cell at the most in a time slot; (c) team of a formation cell has only when all just can be scheduled when empty than the high formation of its priority.Satisfy condition (a) be in order to solve input competition, to avoid sending conflict; Satisfy condition (b) be in order to solve output competition, to avoid receiving conflict; Satisfy condition (c) be in order to guarantee the priority scheduling rule.
According to current research and operating position, support the crossbar dispatching algorithm of multipriority that p-iSLIP is arranged
[5]And OSP
[6]The former is applied to Cisco12000 series router (supporting 32 OC-192 interfaces at most), and the latter is applied to " milky way Alioth " series router (supporting 16 OC-48 interfaces at most) of China Datang Telecom.These two kinds of algorithms exist all that the control information amount is big, the shortcoming of poor expandability, and it is long that this has just limited their application, particularly OSP algorithmic statement time in high speed, multiport, big capacity router, is difficult to satisfy high speed crossbar more and dispatches needs.P-iSLIP and OSP adopt iterative strategy, and each implementation comprises repeatedly iteration, and each iteration comprises " request-permission-acceptance " three phases.With p-iSLIP is that example is simply introduced the running of " request-permission-acceptance ":
Stage 1: request.Each input i selects each VOQ
jThe non-NULL subqueue of middle limit priority sends request to output j, carries precedence information in the request.
Stage 2: permission.If output j receives a plurality of requests, so at first find out the request of all limit priorities, therefrom select one according to the round-robin rule then, and send enabling signal to the input that sends this request.
Stage 3: accept.If input i receives a plurality of permissions, so at first find out the permission of all limit priorities, therefrom select one according to the round-robin rule then, and send acknowledge(ment) signal to the output that sends this permission, so just set up a connection.
For a crossbar who supports the NxN of P priority, table 1 has compared p-iSLIP and OSP algorithm, and wherein control information amount refers in iteration, the amount of information that exchanges between input and the scheduler.
The comparison of table 1p-iSLIP and OSP algorithm
Consider a 64x64 switching network of supporting 8 priority, all of the port speed is identical, is 10Gbps.The cell transmission time of one 64 byte length is 51.2ns, that is to say, implementation of scheduler must be limited in the 51.2ns to be finished.According to table 1, each iteration, the control information of the total 262bit of p-iSLIP and OSP, these transmission of Information will limit the expansion of p-iSLIP and OSP and may reduce its delay performance.According to present technology, the serial electronic circuit is difficult to reach 10Gpbs, even like this, the transmission of these control informations also needs the time of 26.2ns, in fact allows the real time of carrying out of scheduler to have only 51.2-26.2=25ns.These times can't be satisfied the needs of next iteration, and this will cause the decline of delay performance.In addition, the OSP convergence needs iteration N time, limits its application in the multiport router more.Along with port number, priority number purpose increase and the raising of port speed, these situations worsen more.
Development along with router technology, high speed router of future generation is positioned over different frames with crosspoint (scheduler and crossbar) respectively with the input/output interface card gradually, and use optical exchange structure to replace electronics crossbar, to support dozens or even hundreds of high-speed port
[7] [8]Above-mentioned algorithm causes switching network throughput and cell delay performance to descend because the control information amount is big, and the transmission time is long, causes scheduler can't finish scheduling in time in official hour, can not adapt to the development trend of router.
Summary of the invention the invention provides a kind of new support multipriority dispatching method in order to overcome the problems referred to above---p-iDRR (prioritized iterative Dual Round-Robin).The each implementation of p-iDRR also comprises repeatedly iteration, but each iteration only comprises " request-permission " two stages.In " request " stage, each input sends a request at most, and the control information amount is (logN+logP) bit; In " permission " stage, each output sends a permission at the most, and the control information amount is 1bit only.Still with above-mentioned example explanation, the control information amount of the each iteration of p-iDRR has only 10bit, and the transmission time is 1ns.P-iDRR adopts " request-permission " two operation phase, has simplified the hardware realization on the one hand, has improved the speed of service, greatly reduces the control information propagation delay time on the other hand, can satisfy the scheduling needs of high speed, multiport, big capacity router.Compare with similar algorithm, this algorithm has and p-iSLIP, OSP performance much at one, but the hardware realization is simpler; In addition, algorithm the convergence speed is faster than OSP, has overcome their defectives in actual applications.
The technical solution adopted for the present invention to solve the technical problems is:
(1) packet segmentation and reorganization: the IP of different length is grouped in " cell " that is divided into regular length before the forwarding, re-sends to link and get on after the output reorganization.
(2) cell queuing: the cell arrival process is a discrete time random process, and each time slot of each input arrives a cell at the most.VOQ queueing technique is adopted in input rank, is j if arrive a destination at time slot n input i, and priority is that (0≤p≤P-1), this cell is placed into VOQ so for 0≤i, j≤N-1 for the cell of p
j(p+1) individual subqueue in.As long as have in the formation cell etc. to be sent, then can produce the request of transmission.
(3) p-iDRR dispatching algorithm: in the p-iDRR algorithm, each I/O end all has a round-robin moderator, is provided with P pointer, corresponds respectively to P priority.P-iDRR adopts repeatedly the strategy of iteration, and each implementation comprises iteration logN time, and the I/O end that connects in the previous iteration is not participated in the iteration of back, and the iteration of back does not influence the connection of having set up in the previous iteration.All I/O ends were idle condition when each implementation began.The each iteration of p-iDRR includes only two stages:
Stage 1: request.The moderator of input i has P request pointer
Point to the VOQ of current preferential selection respectively.If the request of transmission is arranged among the input i, at first finds out all limit priorities and (be assumed to be p
i) request, moderator uses pointer then
Select a VOQ according to the round-robin rule, and guarantee the corresponding output free time, send request to the output moderator at last, carry precedence information p in the request
i Only upgrade after the iteration in the first time.If the request of being somebody's turn to do is licensed in the stage 2, so
Equal selecteed VOQ port numbers and add 1 (mod N); Otherwise
Equal selecteed VOQ port numbers, should ask simultaneously to remove.
Stage 2: permission.The moderator of output j has P permission pointer g
Jp(0≤p≤P-1,0≤g
Jp≤ N-1), point to the input of current preferential selection respectively.If output j receives request, at first find out all limit priorities and (be assumed to be
) request, moderator uses pointer then
Select the request of an input according to the round-robin rule, and send License Info, thereby set up a connection.
Only upgrade after the iteration, equal selecteed input slogan and add 1 (mod N) in the first time.
(4) the idle informing mechanism of output: in the p-iDRR dispatching algorithm, have following character: in case in front in the iteration certain output receive the request of transmission, then this output must be busy in the iteration of back.According to this character, increase N bar control line, correspond respectively to N output, all input moderators are connected on this N bar control line, in case certain input has request to send to output j, then control line j is changed to high level, represents that this output will be in busy state.All control lines were changed to low level when each implementation began, the expression output free time.Whether all inputs are idle according to the condition judgement output of control line.Utilize the idle informing mechanism of output, p-iDRR can simply determine in the stage that in " request " whether output is idle, do not increase the control information amount between input and the scheduler.
The present invention is further described below in conjunction with drawings and Examples for description of drawings.
Fig. 1 is the composition diagram of the crossbar of input queue switching fabric;
Fig. 2 is an input queuing policy schematic diagram;
Fig. 3 is the implementation of the inventive method one embodiment.
Embodiment is with reference to figure 1, and the crossbar of input queue switching fabric mainly is made up of input rank (VOQ), crossbar, input moderator and centralized scheduler, and all input moderators link to each other with control line.Input rank is used to store and temporarily can not get the cell dispatched; Crossbar is used to set up the connection of I/O end, transmit cell; Input moderator and centralized scheduler are finished the p-iDRR dispatching algorithm jointly, and wherein centralized scheduler mainly is made up of the output moderator.When input has grouping to arrive, at first carry out a series of packet transaction, comprise table look-up, header renewal, classification, segmentation, in input rank, cushion then, wait for the crossbar scheduling.The input moderator is finished the work in " request " stage among the p-iDRR, asks to the scheduler transmission according to the state information of VOQ and control line; Scheduler is finished the work in " permission " stage, makes permission according to solicited message.In addition, the input moderator also is responsible for being provided with the state of control line.All I/O end moderators are separate work.The execution result of scheduler (connection of I/O end) configuration crossbar, thereby the connection of setting up the I/O end, cell is sent to output, is sent to output link through reorganization at last.
Figure 2 shows that the input queuing policy, adopted the VOQ queuing mechanism of supporting multipriority, mainly is the HOL obstructing problem that brings for fear of single FIFO, and input is the formation that each priority of each output is safeguarded an independent FIFO.When specific implementation, these formations can by simple storage management, be divided into the independently formation of a plurality of logics with an independent physical storage.For a crossbar who supports the NxN of P priority, always total (N*P) the individual independent F IFO of each input, the cell process is tabled look-up, classifying is stored in different fifo queues.Priority 0 is limit priority, (P-1) is lowest priority.
Figure 3 shows that the implementation of the inventive method one embodiment.Present embodiment has been showed the process of an iteration of p-iDRR in the crossbar of the 3x3 of 2 priority of a support, and the VOQ formation stain is represented a cell among the figure.Control line is low level when initial, and the state of each pointer as shown in FIG..In " request " stage, it is the request of 0 (high priority) that input 0 and 1 moderator send priority to output 0 simultaneously, has produced the output competition.The output moderator uses g
00Permitted the request of input 1, g
00And r
10Be updated to (1+1) mod 3=2 and (0+1) mod 3=1 respectively; And the request of input 0 does not obtain permission, r
00Be updated to 0, this request simultaneously can be eliminated.Input 0 and 1 moderator can be changed to high level to control line 0 simultaneously.If carry out one time iteration again, though input 0 has the low priority cell to arrive output 0, and low priority pointed VOQ
0, still according to the state of control line, input 0 is busy, and 0 meeting of input is sent the request of low priority to output 1, and obtains permission, finally connects.Since not the iteration first time, r
11And g
11Can not upgrade.
Those skilled in the art are on the present invention program basis, to choose different parameters (iterations, N, P etc.) or to be used for other switching network and other scheme of making, also within the scope of protection of the invention.
[list of references]
[1]Nick?Mckeown?et?al.The?Tiny?Tera:A?packet?switch?core.IEEE?Micro?Magazine,1997,17:26-33.
[2]Cisco?Inc.Cisco?12000?series----Internet?Router.Product?Overview,
http://www.cisco.com,Oct?2001.
[3]Partridge?C.et?al.A?50-Gb/s?IP?router.IEEE?Trans.on?Networking,1998,6:237-248.
[4]M.J.Karol,M.Hluchyj,and?S.Morgan.Input?versus?output?queuing?on?a?space-division?packet?switch.IEEETrans.On?Communications,1987,35:1347-1356.
[5]N.Mckeown.Scheduling?algorithm?for?input-queued?cell?switches.Ph.D.Thesis,UC?Berkeley,May?1995.
[6] Sun Zhigang, Lu Xicheng. a kind of crossbar dispatching algorithm that is used for the Differentiated Services router. National University of Defense technology's journal, 2000,22 (6): 52~56.
[7] Peng Laixian, Li Wanlin, Tian Chang, Zheng Shaoren. too bit router key technology is analyzed. " telecommunications science ", 2002,18 (3): 11~15.
[8]A?New?architecture?for?switch?and?router?design,PMC-Sierra,Inc.White?paper,1999.
Claims (5)
1, a kind of high speed crossbar dispatching method of supporting multipriority comprises queueing technique and crossbar dispatching algorithm, it is characterized in that:
(1), the IP of different length is grouped in " cell " that is divided into regular length before the forwarding, after the output reorganization, re-send to link and get on; Grouping is only stored at input, virtual output work queue technology (VOQ) is adopted in the input formation, for each priority of each output is safeguarded an independently first in first out (FIFO) formation, if a destination is that j, priority are the cell arrival input i of p, 0≤i wherein, j≤N-1,0≤p≤P-1, this cell is placed into VOQ so
jP+1 subqueue in;
(2), the crossbar dispatching algorithm is called p-iDRR, in this algorithm, each I/O end all has a round-robin moderator, is provided with P pointer, corresponds respectively to P priority; The each iteration of p-iDRR comprises " request " and " permission " two stages:
Stage 1: request; The moderator of input i has P request pointer r
Ip, 0≤i≤N-1 wherein, 0≤p≤P-1,0≤r
Ip≤ N-1 points to the VOQ of current preferential selection respectively, if the request of transmission is arranged among the input i, at first finding out all limit priorities is p
iRequest, moderator uses pointer then
Select a VOQ according to the round-robin rule, and guarantee the corresponding output free time, send request to the output moderator at last, carry precedence information p in the request
i Only upgrade after the iteration, wherein in the first time
If the request of being somebody's turn to do is licensed in the stage 2, so
Equal selecteed VOQ port numbers and add 1 modN as a result; Otherwise
Equal selecteed VOQ port numbers, should ask simultaneously to remove;
Stage 2: permission; The moderator of output j has P permission pointer g
Jp, 0≤j≤N-1 wherein, 0≤p≤P-1,0≤g
Ip≤ N-1 points to the input of current preferential selection respectively, if output j receives request, at first finds out all limit priorities and is
Request, moderator uses pointer then
Wherein
Select the request of an input according to the round-robin rule, and send License Info, thereby set up a connection;
Only upgrade after the iteration, equal selecteed input slogan and add 1 the N of mod as a result in the first time.
2, crossbar dispatching method as claimed in claim 1, it is characterized in that: the idle informing mechanism of output is connected in N bar control line with all input moderators, correspond respectively to N output, in case certain input has request to send to output j, 0≤j≤N-1 wherein, then control line j is changed to high level, represents that this output will be non-idle condition.
3, crossbar dispatching method as claimed in claim 1, it is characterized in that: the strategy that adopts repeatedly iteration, each implementation comprises iteration logN time, wherein N is an input/outbound port number, the I/O end that connects in the previous iteration is not participated in the iteration of back, and the iteration of back does not influence the connection of having set up in the previous iteration; All I/O ends were idle condition when each implementation began.
4, crossbar dispatching method as claimed in claim 1, it is characterized in that: each iteration comprises " request " and " permission " two stages, in " request " stage, each input sends a request at the most, and the control information amount is (logN+logP) bit; In " permission " stage, each output sends a permission at the most, and the control information amount is 1 bit only.
5, crossbar dispatching method as claimed in claim 1 is characterized in that: the cell size of crossbar exchange is 64 bytes, and port speed is 10Gbps.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CNB031583903A CN100490383C (en) | 2003-09-29 | 2003-09-29 | A high-speed Crossbar scheduling method for supporting multipriority |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CNB031583903A CN100490383C (en) | 2003-09-29 | 2003-09-29 | A high-speed Crossbar scheduling method for supporting multipriority |
Publications (2)
Publication Number | Publication Date |
---|---|
CN1604537A CN1604537A (en) | 2005-04-06 |
CN100490383C true CN100490383C (en) | 2009-05-20 |
Family
ID=34660455
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CNB031583903A Expired - Fee Related CN100490383C (en) | 2003-09-29 | 2003-09-29 | A high-speed Crossbar scheduling method for supporting multipriority |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN100490383C (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108173787A (en) * | 2017-11-30 | 2018-06-15 | 成都成电光信科技股份有限公司 | Data transmission path selection method, Method of Scheduling Parallel and the device of FC interchangers |
Families Citing this family (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN100466602C (en) * | 2005-04-28 | 2009-03-04 | 华为技术有限公司 | Method for dynamically sharing space of memory |
CN1870769B (en) * | 2005-05-26 | 2011-12-07 | 华为技术有限公司 | Arbitration method and unit of cross connection matrix based on priority propagation |
CN100421420C (en) * | 2005-08-02 | 2008-09-24 | 华为技术有限公司 | Method for dispatching variable length data packet queue in crossbar switching matrix |
CN100373892C (en) * | 2005-09-13 | 2008-03-05 | 中兴通讯股份有限公司 | Virtual exchanging method capable of routing |
CN100550832C (en) * | 2005-11-15 | 2009-10-14 | 华为技术有限公司 | A kind of arbitration implementation method of switching network |
CN101383753B (en) * | 2007-09-04 | 2010-12-29 | 普然通讯技术(上海)有限公司 | Method for realizing VOQ scheduling in device cascade system |
CN103023669B (en) * | 2012-12-28 | 2015-04-15 | 中国航空工业集团公司第六三一研究所 | Switching network based broadcast scheduling method |
CN104104617B (en) * | 2014-08-07 | 2017-10-17 | 曙光信息产业(北京)有限公司 | A kind of message referee method and device |
CN105429898B (en) * | 2015-11-02 | 2018-04-06 | 北京航空航天大学 | A kind of CICQ structures intersect the balanced packet scheduling algorithm of buffer queue |
CN105429909B (en) * | 2015-12-31 | 2018-11-20 | 上海交通大学 | A kind of parallel switching dispatching method based on secondary color |
CN108234360A (en) * | 2016-12-14 | 2018-06-29 | 中国航空工业集团公司西安航空计算技术研究所 | A kind of antiblocking FC switch concatenations port is realized and method |
CN107171973B (en) * | 2017-05-04 | 2020-06-09 | 内蒙古农业大学 | Two-stage exchange structure implementation method based on adjacent port scheduling information |
CN107360107A (en) * | 2017-08-25 | 2017-11-17 | 东南大学 | A kind of cross bar switch dispatching method of the low delay suitable for network-on-chip |
CN111711574B (en) * | 2020-06-29 | 2022-05-24 | 中国人民解放军国防科技大学 | Ultra-high order single-cycle message scheduling method and device |
CN115150343B (en) * | 2022-09-05 | 2023-01-03 | 井芯微电子技术(天津)有限公司 | Exchange chip verification platform device and method based on UVM |
-
2003
- 2003-09-29 CN CNB031583903A patent/CN100490383C/en not_active Expired - Fee Related
Non-Patent Citations (2)
Title |
---|
基于螺旋线的Round-Robin Crossbar调度算法. 张志群,魏激波,丁炜.电子与信息学报,第25卷第6期. 2003 |
基于螺旋线的Round-Robin Crossbar调度算法. 张志群,魏激波,丁炜.电子与信息学报,第25卷第6期. 2003 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108173787A (en) * | 2017-11-30 | 2018-06-15 | 成都成电光信科技股份有限公司 | Data transmission path selection method, Method of Scheduling Parallel and the device of FC interchangers |
CN108173787B (en) * | 2017-11-30 | 2020-12-18 | 成都成电光信科技股份有限公司 | Data transmission path selection method, parallel scheduling method and device for FC switch |
Also Published As
Publication number | Publication date |
---|---|
CN1604537A (en) | 2005-04-06 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US10182021B2 (en) | Crossbar switch and recursive scheduling | |
Chuang et al. | Practical algorithms for performance guarantees in buffered crossbars | |
CN100490383C (en) | A high-speed Crossbar scheduling method for supporting multipriority | |
Iyer et al. | Analysis of the parallel packet switch architecture | |
KR100339329B1 (en) | RRGS-Round-Robin Greedy Scheduling for input/output terabit switches | |
US7023840B2 (en) | Multiserver scheduling system and method for a fast switching element | |
USRE43110E1 (en) | Pipelined maximal-sized matching cell dispatch scheduling | |
US7046661B2 (en) | Scheduling the dispatch of cells in non-empty virtual output queues of multistage switches using a pipelined hierarchical arbitration scheme | |
US20060285548A1 (en) | Matching process | |
KR19990087752A (en) | Efficient Output Request Packet Switch and Method | |
US7203202B2 (en) | Arbitration using dual round robin matching with exhaustive service of winning virtual output queue | |
Shen et al. | Byte-focal: A practical load balanced switch | |
US6865154B1 (en) | Method and apparatus for providing bandwidth and delay guarantees in combined input-output buffered crossbar switches that implement work-conserving arbitration algorithms | |
Smiljanic et al. | RRGS-round-robin greedy scheduling for electronic/optical terabit switches | |
US7079545B1 (en) | System and method for simultaneous deficit round robin prioritization | |
Elhanany et al. | Packet scheduling in next-generation multiterabit networks | |
Obara et al. | High speed transport processor for broad-band burst transport system | |
Blanton et al. | Impact of polarized traffic on scheduling algorithms for high-speed optical switches | |
Kleban et al. | Packet dispatching algorithms with the static connection patterns scheme for three-stage buffered clos-network switches | |
Zheng et al. | An efficient round-robin algorithm for combined input-crosspoint-queued switches | |
Yun | A terabit multi-service switch with Quality of Service support | |
Leonardi et al. | Algorithms for virtual output queued switching | |
Guo et al. | Packet switch with internally buffered crossbars | |
KR100368439B1 (en) | Method and apparatus for ensuring transfer order in a packet-switch with a dual-switching plane000 | |
Jeong et al. | An advanced input-queued ATM switch with a pipelined approach to arbitration |
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: 20090520 Termination date: 20091029 |