Dodging the cost of unavoidable memory copies in message logging protocols

Authors:
George Bosilca;Aurelien Bouteiller;Thomas Herault;Pierre Lemarinier;Jack J. Dongarra
Affiliations:
University of Tennessee, TN;University of Tennessee, TN;University of Tennessee, TN and Universite Paris-Sud, INRIA, France;University of Tennessee, TN;University of Tennessee, TN and Oak Ridge National Laboratory, TN
Venue:
EuroMPI'10 Proceedings of the 17th European MPI users' group meeting conference on Recent advances in the message passing interface
Year:
2010

Citing 7
Cited 2

MPI: a message passing interface

Proceedings of the 1993 ACM/IEEE conference on Supercomputing
Optimizing memory system performance for communication in parallel computers

ISCA '95 Proceedings of the 22nd annual international symposium on Computer architecture
OPIOM: off-processor I/O with myrinet

Future Generation Computer Systems - Best papers from symp. on cluster computing and the grid (CCGRID 2001)
A low-overhead recovery technique using quasi-synchronous checkpointing

ICDCS '96 Proceedings of the 16th International Conference on Distributed Computing Systems (ICDCS '96)
Improved message logging versus improved coordinated checkpointing for fault tolerant MPI

CLUSTER '04 Proceedings of the 2004 IEEE International Conference on Cluster Computing
Efficient asynchronous memory copy operations on multi-core systems and I/OAT

CLUSTER '07 Proceedings of the 2007 IEEE International Conference on Cluster Computing
Retrospect: deterministic replay of MPI applications for interactive distributed debugging

PVM/MPI'07 Proceedings of the 14th European conference on Recent Advances in Parallel Virtual Machine and Message Passing Interface

Correlated set coordination in fault tolerant message logging protocols

Euro-Par'11 Proceedings of the 17th international conference on Parallel processing - Volume Part II
Alleviating scalability issues of checkpointing protocols

SC '12 Proceedings of the International Conference on High Performance Computing, Networking, Storage and Analysis

Quantified Score

Hi-index	0.00

Visualization

Abstract

With the number of computing elements spiraling to hundred of thousands in modern HPC systems, failures are common events. Few applications are nevertheless fault tolerant; most are in need for a seamless recovery framework. Among the automatic fault tolerant techniques proposed for MPI, message logging is preferable for its scalable recovery. The major challenge for message logging protocols is the performance penalty on communications during failure-free periods, mostly coming from the payload copy introduced for each message. In this paper, we investigate different approaches for logging payload and compare their impact on network performance.