Characterization of deadlocks in interconnection networks

Warnakulasuriya, Sugath; Pinkston, Timothy M.

doi:10.1109/ipps.1997.580852

Cited by 26 publications

(26 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Thus, routing algorithms usually have some constraints in order to avoid deadlocks. Recently, the frequency of deadlock occurrence in k-ary n-cube networks using wormhole switching was measured emperically [21,18]. From this study, we know that deadlocks are very rare, especially when two or more virtual channels are used with true fully adaptive routing.…”

Section: Software-based Deadlock Recoverymentioning

confidence: 95%

“…Provided that there are some escape resources to drain messages blocking cyclically, deadlock cannot occur, but messages wait for a long time in the network. One solution to this problem is to increase routing freedom by adding more virtual channels [9,21,18]; then messages will have less probability of being involved in cyclic dependencies. However, an excessive number of virtual channels could lead to a reduced clock frequency [6].…”

Section: Message Injection Limitationmentioning

confidence: 99%

“…However, both techniques may produce severe performance degradation when the network is close to saturation. Performance degradation at the saturation point was studied in [9,15] and, more recently, in [21,18]. In [9], this situation was described as occurring when messages block cyclically faster than they are drained using the escape path.…”

Section: Introductionmentioning

confidence: 99%

“…Although this can be mitigated by adding several virtual channels per physical channel, this solution is expensive and is overkill for most networks. In [21,18], the frequency of deadlock occurrence on k-ary n-cubes using a true fully adaptive minimal routing algorithm with deadlock recovery was measured. It was shown that deadlocks rarely occur when sufficient routing freedom is provided, but they are more likely to occur when the network is close to or beyond saturation.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Software-based deadlock recovery technique for true fully adaptive routing in wormhole networks

Martínez

López²,

Duato³

et al.

Proceedings of the 1997 International Conference on Parallel Processing (Cat. No.97TB100162)

Self Cite

View full text Add to dashboard Cite

show abstract

Section: Software-based Deadlock Recoverymentioning

confidence: 95%

Section: Message Injection Limitationmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Software-based deadlock recovery technique for true fully adaptive routing in wormhole networks

Martínez

López²,

Duato³

et al.

Proceedings of the 1997 International Conference on Parallel Processing (Cat. No.97TB100162)

Self Cite

View full text Add to dashboard Cite

show abstract

“…In either deadlock avoidance or recovery, the frequency of deadlocks in the adaptive channels increases dramatically when the network reaches saturation [33]. When this occurs, packets are delivered over the relatively limited escape bandwidth available on the deadlock-free paths.…”

Section: Background and Related Workmentioning

confidence: 99%

Exploiting global knowledge to achieve self-tuned congestion control for k-ary n-cube networks

Thottethodi

Lebeck

Mukherjee

2004

IEEE Trans. Parallel Distrib. Syst.

View full text Add to dashboard Cite

Abstract-Network performance in tightly-coupled multiprocessors typically degrades rapidly beyond network saturation. Consequently, designers must keep a network below its saturation point by reducing the load on the network. Congestion control via source throttling-a common technique to reduce the network load-prevents new packets from entering the network in the presence of congestion. Unfortunately, prior schemes to implement source throttling either lack vital global information about the network to make the correct decision (whether to throttle or not) or depend on specific network parameters, or communication patterns. This paper presents a global-knowledge-based, self-tuned, congestion control technique that prevents saturation at high loads across different communication patterns for k-ary n-cube networks. Our design is composed of two key components. First, we use global information about a network to obtain a timely estimate of network congestion. We compare this estimate to a threshold value to determine when to throttle packet injection. The second component is a self-tuning mechanism that automatically determines appropriate threshold values based on throughput feedback. A combination of these two techniques provides high performance under heavy load, does not penalize performance under light load, and gracefully adapts to changes in communication patterns.

show abstract