Memory bandwidth aware scheduling for SMP cluster nodes

Koukis, E; Koziris, N

dc.contributor.author	Koukis, E	en
dc.contributor.author	Koziris, N	en
dc.date.accessioned	2014-03-01T02:43:25Z
dc.date.available	2014-03-01T02:43:25Z
dc.date.issued	2005	en
dc.identifier.uri	https://dspace.lib.ntua.gr/xmlui/handle/123456789/31390
dc.subject	Continuous Improvement	en
dc.subject	Design and Implementation	en
dc.subject	Memory Bandwidth	en
dc.subject	Performance Monitoring	en
dc.subject	Processor Sharing	en
dc.subject	Scheduling Algorithm	en
dc.subject	Shared Memory	en
dc.subject	Smp Cluster	en
dc.subject	Time Use	en
dc.subject.other	Algorithms	en
dc.subject.other	Bandwidth	en
dc.subject.other	Computer aided design	en
dc.subject.other	Computer architecture	en
dc.subject.other	Computer networks	en
dc.subject.other	Program processors	en
dc.subject.other	Memory bandwidth	en
dc.subject.other	Multiprogrammed workloads	en
dc.subject.other	Scheduling algorithm	en
dc.subject.other	SMP cluster nodes	en
dc.subject.other	System processors	en
dc.subject.other	Storage allocation (computer)	en
dc.title	Memory bandwidth aware scheduling for SMP cluster nodes	en
heal.type	conferenceItem	en
heal.identifier.primary	10.1109/EMPDP.2005.26	en
heal.identifier.secondary	http://dx.doi.org/10.1109/EMPDP.2005.26	en
heal.identifier.secondary	1386058	en
heal.publicationDate	2005	en
heal.abstract	Clusters of SMPs are becoming increasingly common. However, the shared memory design of SMPs and the consequential contention between system processors for access to main memory can limit their efficiency significantly. Moreover, the continuous improvement of modern cluster interconnection technologies leads to the network bandwidth being a significant fraction of the total memory bandwidth of the machine, thus the NIC of an SMP cluster node can also become a major consumer of shared memory bus bandwidth. In this paper we first provide experimental evidence that contention on the shared memory bus can have major impact on the total execution time of processes even when no processor sharing is involved, then present the design and implementation of an informed scheduling algorithm for multiprogrammed workloads, which tries to carefully select processes to be co-scheduled so that bus saturation is avoided. The input data needed by our scheduler are acquired dynamically, at run-time, using architecture-specific performance monitoring counters and a modified version of the NIC firmware, with no changes to existing application binaries. Experimental comparison between our scheduler and the standard Linux 2.6 O(1) scheduler shows average system throughput improvements in the range of 5-25%.	en
heal.journalName	Proceedings - 13th Euromicro Conference on Parallel, Distributed and Network-Based Processing 2005, PDP 2005	en
dc.identifier.doi	10.1109/EMPDP.2005.26	en
dc.identifier.volume	2005	en
dc.identifier.spage	187	en
dc.identifier.epage	196	en