2002
DOI: 10.1109/tnet.2002.804823
|View full text |Cite
|
Sign up to set email alerts
|

Concurrent round-robin-based dispatching schemes for Clos-network switches

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
1
1

Citation Types

0
106
0
1

Year Published

2004
2004
2019
2019

Publication Types

Select...
5
2
1

Relationship

1
7

Authors

Journals

citations
Cited by 104 publications
(107 citation statements)
references
References 15 publications
0
106
0
1
Order By: Relevance
“…For heavy input loads ( p > 0.75), the aver- 2 and gr. 4), published in [13], with the results obtained for the MCNS, SDRUB, pure ML MWM DD, and CRRD (concurrent round-robin dispatching) [9]. It is possible to see that the ML DISQUO DD algorithm performs poorly, but the MCNS produces the shortest average cell delay using a simpler packet dispatching scheme.…”
Section: Performance Evaluationmentioning
confidence: 88%
See 1 more Smart Citation
“…For heavy input loads ( p > 0.75), the aver- 2 and gr. 4), published in [13], with the results obtained for the MCNS, SDRUB, pure ML MWM DD, and CRRD (concurrent round-robin dispatching) [9]. It is possible to see that the ML DISQUO DD algorithm performs poorly, but the MCNS produces the shortest average cell delay using a simpler packet dispatching scheme.…”
Section: Performance Evaluationmentioning
confidence: 88%
“…Most of the packet dispatching schemes developed for the three-stage Clos-network switches insist on using the request-grant-accept handshaking routine, based on the effect of the desynchronization of arbitration pointers [9,10]. In this routine, if cells are waiting for transmission in queues, input arbiters send requests to corresponding output port arbiters.…”
Section: Related Work and Motivationmentioning
confidence: 99%
“…We evaluate RD, CRRD, and MOMD under this traffic pattern. To obtain the best performance, CRRD adopts iIM = m (i.e., the number of iterations performed in IM to match the VOQs used in CRRD and output links [5]) for any number iIM -CM . Figure 5 shows that RD, with one iteration, and CRRD, with multiple iIM -CM iterations, cannot offer 100% throughput under unbalanced traffic.…”
Section: Performance Evaluation Of Momdmentioning
confidence: 99%
“…Moreover, diverse dispatching schemes have been proposed to reduce the average cell delay [5], [11] under uniform traffic. However, real traffic patterns are not only uniform, but a wide variety of admissible traffic patterns, including those with nonuniform distributions.…”
Section: Introductionmentioning
confidence: 99%
“…As for the commonly studied three-stage Clos-network; it can be a Space-SpaceSpace (S3) network without buffers or Memory-MemoryMemory (MMM) [3]- [6] with buffered switching units at all stages. Other combinations have also been studied [7]- [10] by virtue of achieving good performance for less complex hardware and scheduling. Despite their scalability potential, almost all existing Clos-network based proposals are either subject to considerable implementation complexity, prohibitively high cost or poor performance.…”
mentioning
confidence: 99%