GMBlock: Optimizing data movement in a block-level storage sharing system over Myrinet

Koukis, E; Nanos, A; Koziris, N

dc.contributor.author	Koukis, E	en
dc.contributor.author	Nanos, A	en
dc.contributor.author	Koziris, N	en
dc.date.accessioned	2014-03-01T01:33:34Z
dc.date.available	2014-03-01T01:33:34Z
dc.date.issued	2010	en
dc.identifier.issn	1386-7857	en
dc.identifier.uri	https://dspace.lib.ntua.gr/xmlui/handle/123456789/20468
dc.subject	Block-level storage	en
dc.subject	Memory contention	en
dc.subject	Myrinet	en
dc.subject	Network block device	en
dc.subject	OCFS2	en
dc.subject	Shared storage	en
dc.subject	SMP clusters	en
dc.subject	User level networking	en
dc.subject.classification	Computer Science, Information Systems	en
dc.subject.classification	Computer Science, Theory & Methods	en
dc.subject.other	Block-level storage	en
dc.subject.other	Memory contentions	en
dc.subject.other	Myrinet	en
dc.subject.other	Network block device	en
dc.subject.other	OCFS2	en
dc.subject.other	Shared storage	en
dc.subject.other	SMP clusters	en
dc.subject.other	User-level networking	en
dc.subject.other	Benchmarking	en
dc.subject.other	Optimization	en
dc.subject.other	Disks (structural components)	en
dc.title	GMBlock: Optimizing data movement in a block-level storage sharing system over Myrinet	en
heal.type	journalArticle	en
heal.identifier.primary	10.1007/s10586-009-0106-y	en
heal.identifier.secondary	http://dx.doi.org/10.1007/s10586-009-0106-y	en
heal.language	English	en
heal.publicationDate	2010	en
heal.abstract	We present gmblock, a block-level storage sharing system over Myrinet which uses an optimized I/O path to transfer data directly between the storage medium and the network, bypassing the host CPU and main memory bus of the storage server. It is device driver independent and retains the protection and isolation features of the OS. We evaluate the performance of a prototype gmblock server and find that: (a) the proposed techniques eliminate memory and peripheral bus contention, increasing remote I/O bandwidth significantly, in the order of 20-200% compared to an RDMA-based approach, (b) the impact of remote I/O to local computation becomes negligible, (c) the performance characteristics of RAID storage combined with limited NIC resources reduce performance. We introduce synchronized send operations to improve the degree of disk to network I/O overlapping. We deploy the OCFS2 shared-disk filesystem over gmblock and show gains for various application benchmarks, provided I/O scheduling can eliminate the disk bottleneck due to concurrent access. © 2010 Springer Science+Business Media, LLC.	en
heal.publisher	SPRINGER	en
heal.journalName	Cluster Computing	en
dc.identifier.doi	10.1007/s10586-009-0106-y	en
dc.identifier.isi	ISI:000284300700001	en
dc.identifier.volume	13	en
dc.identifier.issue	4	en
dc.identifier.spage	349	en
dc.identifier.epage	372	en