Efficient Communication Using Message Prediction for Cluster of Multiprocessors

Afsahi, Ahmad; Dimopoulos, N.J.

doi:10.1007/10720115_12

Cited by 16 publications

(28 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Communication latencies have been identified as one of the performance limiting factors of message passing applications in clusters of workstations/multiprocessors [1]. On the receiver side, message-copying operations contribute to these communication latencies.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Predicting MPI Buffer Addresses

Freitag¹,

Farreras²,

Cortés³

et al. 2004

Computational Science - ICCS 2004

View full text Add to dashboard Cite

Abstract. Communication latencies have been identified as one of the performance limiting factors of message passing applications in clusters of workstations/multiprocessors. On the receiver side, message-copying operations contribute to these communication latencies. Recently, prediction of MPI messages has been proposed as part of the design of a zero message-copying mechanism. Until now, prediction was only evaluated for the next message. Predicting only the next message, however, may not be enough for real implementations, since messages do not arrive in the same order as they are requested. In this paper, we explore long-term prediction of MPI messages for the design of a zero message-copying mechanism. To achieve long-term prediction we evaluate two prediction schemes, the first based on graphs, and the second based on periodicity detection. Our experiments indicate that with both prediction schemes the buffer addresses and message sizes of several future MPI messages (up to +10) can be predicted successfully.

show abstract

Section: Introductionmentioning

confidence: 99%

“…Zero message-copying on the receiver side of MPI has been indicated as a technique to reduce this communication latency [1]. One of the identified requirements to achieve zero message-copying is to predict the characteristics of the next MPI message.…”

Section: Introductionmentioning

confidence: 99%

Predicting MPI Buffer Addresses

Freitag¹,

Farreras²,

Cortés³

et al. 2004

Computational Science - ICCS 2004

View full text Add to dashboard Cite

show abstract

“…In the third, and fourth column the number of point-to-point and collective messages received by a process is shown. The remaining columns indicate the number of different message sizes and different sender processes appearing in the MPI message stream received by a process 1 . In Table 1 it can be observed that the message stream of these benchmarks typically is of a few thousand messages, with exception of the LU benchmark, which has tens of thousand of messages.…”

Section: Benchmarks Usedmentioning

confidence: 99%

“…In [1,2] the prediction of MPI messages are proposed. In this work a number of heuristics for the prediction of MPI messages are presented.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Exploring the predictability of MPI messages

Freitag¹,

Caubet²,

Farrera³

et al.

Proceedings International Parallel and Distributed Processing Symposium

View full text Add to dashboard Cite

show abstract

Efficient communication using message prediction for clusters of multiprocessors

Afsahi

Dimopoulos

2002

Concurrency and Computation

Self Cite

View full text Add to dashboard Cite

SUMMARYWith the increasing uniprocessor and symmetric multiprocessor computational power available today, interprocessor communication has become an important factor that limits the performance of clusters of workstations/multiprocessors. Many factors including communication hardware overhead, communication software overhead, and the user environment overhead (multithreading, multiuser) affect the performance of the communication subsystems in such systems. A significant portion of the software communication overhead belongs to a number of message copying operations. Ideally, it is desirable to have a true zerocopy protocol where the message is moved directly from the send buffer in its user space to the receive buffer in the destination without any intermediate buffering. However, due to the fact that message-passing applications at the send side do not know the final receive buffer addresses, early arrival messages have to be buffered at a temporary area. In this paper, we show that there is a message reception communication locality in message-passing applications. We have utilized this communication locality and devised different message predictors at the receiver sides of communications. In essence, these message predictors can be efficiently used to drain the network and cache the incoming messages even if the corresponding receive calls have not yet been posted. The performance of these predictors, in terms of hit ratio, on some parallel applications are quite promising and suggest that prediction has the potential to eliminate most of the remaining message copies. We also show that the proposed predictors do not have sensitivity to the starting message reception call, and that they perform better than (or at least equal to) our previously proposed predictors.

show abstract

Efficient Communication Using Message Prediction for Cluster of Multiprocessors

Cited by 16 publications

References 27 publications

Predicting MPI Buffer Addresses

Predicting MPI Buffer Addresses

Exploring the predictability of MPI messages

Efficient communication using message prediction for clusters of multiprocessors

Contact Info

Product

Resources

About