[Colloquium] Barati/MS Presentation/Nov 24, 2015

[Margaret Jaffey]

This is an announcement of Saeid Barati's MS Presentation.

------------------------------------------------------------------------------
Date:  Tuesday, November 24, 2015

Time:  3:00 PM

Place:  Ryerson 255

M.S. Candidate:  Saeid Barati

M.S. Paper Title: PROVIDING FAIRNESS IN HETEROGENEOUS MULTICORES WITH
A PREDICTIVE, ADAPTIVE SCHEDULER

Abstract:
On multicore processors applications contend for resources –
especially memory bandwidth – re- ducing both fairness and overall
system performance. To address this problem, contention-aware
schedulers have been proposed to provide fairness and predictable
behavior through resource man- agement. These approaches have two
limitations: 1) Many introduce overhead that reduces per- formance. 2)
The emergence of heterogeneous multicores has made the handling
contention and providing fairness much more difficult. This thesis
made two major contributions. First we present Simception, a fast and
flexible mul- ticore simulator that can be used to design and
experiment different contention-aware scheduling policies. We verify
Simception with performance output of TilePro64 manycore system.
Further- more, we propose augmenting existing contention-aware
approaches with lightweight predictive and adaptive components named
as Dike to provide fair memory access and performance improve- ments
on heterogeneous multicores. The predictive component’s closed-loop
approach anticipates how different processes will perform with
different core types, while the adaptive component dy- namically tunes
key scheduling parameters to the current workload. We implement and
evaluate this approach on a real Linux/x86 system with a variety of
mem- ory and compute intensive benchmarks. We find that adding
prediction improves fairness and performance by 30% and 4%
(respectively) compared to a prior state-of-the-art contention-aware
approach. The addition of adaptation allows users to select for
fairness or performance optimiza- tion, providing an additional 24%
improvement in fairness or a 9% improvement in performance beyond the
predictive approach.

Saeid's advisor is Prof. Henry Hoffmann

Login to the Computer Science Department website for details:
 https://www.cs.uchicago.edu/phd/ms_announcements#saeid

=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
Margaret P. Jaffey            margaret at cs.uchicago.edu
Department of Computer Science
Student Support Rep (Ry 156)               (773) 702-6011
The University of Chicago      http://www.cs.uchicago.edu
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=