Home  |   Login  |   Logout  |   Access Information  |   Alerts  |   Purchase History  |   Cart  |   Sitemap  |   Help   
 
Abstract
BROWSE SEARCH IEEE XPLORE GUIDE SUPPORT
arrow_left View Search Results  
Email/Printer Friendly Format  
 

A comparison of local and gang scheduling on a Beowulf cluster

Strazdins, P.   Uhlmann, J.  
Dept. of Comput. Sci., Australian Nat. Univ., Acton, ACT, Australia
This paper appears in: Cluster Computing, 2004 IEEE International Conference on
Publication Date: 20-23 Sept. 2004
On page(s): 55 - 62
ISSN: 1552-5244
ISBN: 0-7803-8694-9
Digital Object Identifier: 10.1109/CLUSTR.2004.1392601
Current Version Published: 2005-02-22

Abstract
Gang scheduling and related techniques are widely believed to be necessary for efficient job scheduling on distributed memory parallel computers. This is because they minimize context switching overheads and permit the parallel job currently running to progress at the fastest possible rate. However, in the case of cluster computers, and particularly those with COTS networks, these benefits can be outweighed in the multiple jobs time-sharing context by the loss the ability to utilize the CPU for other jobs when the current job is waiting for messages. Experiments on a Linux Beowulf cluster with 100 Mb fast Ethernet switches are made comparing the SCore buddy-based gang scheduling with local scheduling (provided by the Linux 2.4 kernel with MPI implemented over TCP/IP). Results for communication-intensive numerical applications on 16 nodes reveal that gang scheduling results in 'slowdowns ' up to a factor of two greater for 8 simultaneous jobs. This phenomenon is not due to any deficiencies in SCore but due to the relative costs of context switching versus message overhead, and we expect similar results holds for any gang scheduling implementation. A performance analysis of local scheduling indicates that cache pollution due to context switching is more significant than the direct context switching overhead on the applications studied. When this is taken into account, local scheduling behaviour comes close to achieving ideal slowdowns for finer-grained computations such as Linpack. The performance models also indicate that similar trends are to be expected for clusters with faster networks.

Index Terms
Available to subscribers and IEEE members.

References
Available to subscribers and IEEE members.
Citing Documents
Available to subscribers and IEEE members.
You are not logged in.
Guests may access Abstract records free of charge.
Login
Username
Password
» Forgot your password?
Please remember to log out when you have finished your session.
You must log in to access:
• Advanced or Author Search
• CrossRef Search
• AbstractPlus Records
• Full Text PDF
• Full Text HTML
Access this document
Full Text PDF icon
Full Text: PDF (779 KB)
» Buy this document now
» Learn more about
» Learn more about
   purchasing articles
   and standards
Rights and Permissions>
» Learn More
Download this citation
Available to subscribers and IEEE members.
 
arrow_left View Search Results  
Indexed by IEE Inspec
© Copyright 2010 IEEE – All Rights Reserved