By Topic

The Performance of Erasure Codes Used in FT-MPI

Sign In

Cookies must be enabled to login.After enabling cookies , please use refresh or reload or ctrl+f5 on the browser for the login options.

Formats Non-Member Member
$33 $13
Learn how you can qualify for the best price for this item!
Become an IEEE Member or Subscribe to
IEEE Xplore for exclusive pricing!
close button

puzzle piece

IEEE membership options for an individual and IEEE Xplore subscriptions for an organization offer the most affordable access to essential journal articles, conference papers, standards, eBooks, and eLearning courses.

Learn more about:

IEEE membership

IEEE Xplore subscriptions

5 Author(s)
Liu Xiaoguang ; Nankai-Baidu Joint Lab., Nankai Univ., Tianjin, China ; Wang Gang ; Zhang Yu ; Li Ang
more authors

Today, the scale of high performance computing (HPC) systems is much larger than ever. Some HPC systems consist of thousands or even tens of thousands of processors. The larger scale leads to a challenge that how to deal with process failures. The most important programming tool for HPC is MPI (message passing interface). There are some existing methods to deal with fault-tolerance, such as MPICH-V, StarFish, MPI/FT and so on, using the MPI context. Most of them do the checkpoint on disk. In this paper, some erasure codes, which used in RAID systems usually, are applied to deal with the fault-tolerance in-memory. Based on fault-tolerance-MPI (FT-MPI) platform, RAID4, RAID5, RDP and X-code are implanted to do the checkpoint in-memory. The experimental results show that RDP is feasible for double-fault-tolerance in-memory.

Published in:

Information Technology and Applications, 2009. IFITA '09. International Forum on  (Volume:3 )

Date of Conference:

15-17 May 2009