[LA-SiGMA-gpu] Supada Laosooksathit (candidate to lead the GPU team) will talk at tomorrow 11am
Ka Ming Tam
phy.kaming at gmail.com
Tue Jun 4 14:35:34 CDT 2013
Dear GPU Group Members,
A candidate for replacing Zhifeng to lead the GPU team, Supada
Laosooksathit,
will give a presentation tomorrow 11am at 338 Johnston.
Please see the following for the title and abstract of her talk.
Best,
Ka-Ming
Time : 11am June 5th (tomorrow).
Place : 338 Johnston
Title : Performance Model for Predicting Run-time of a GPGPU Application
Abstract:
Due to the fact that the reliability of very large scaled systems is
inversely related to the number of computing elements, fault tolerance has
become a major concern in high performance computing (HPC) including the
most recent deployment with GPUs. Many fault tolerance strategies, such as
checkpoint/restart mechanism, have been studied in order to mitigate
failures and the effectiveness.
This talk provides an idea to model techniques that explore interplay
between application performance and system reliability. More importantly,
these two parameters play significant roles toward an optimal outcome in
mitigating faults in a very large system. We taken into account in our
fault tolerance strategies that balance between the application
time-to-completion, and the time-to-failure (TTF) of the system. The former
factor can be estimated by a performance model while the latter can be
approximated by a reliability model.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://mail.loni.org/mailman/private/lasigma-gpu/attachments/20130604/2d3b885a/attachment.html
More information about the LASiGMA-gpu
mailing list