BEGIN:VCALENDAR PRODID:-//Microsoft Corporation//Outlook MIMEDIR//EN VERSION:1.0 BEGIN:VEVENT DTSTART:20111116T193000Z DTEND:20111116T200000Z LOCATION:TCC 304 DESCRIPTION;ENCODING=QUOTED-PRINTABLE:ABSTRACT: Infrastructure-as-a-Service (IaaS) cloud computing is gaining=0Asignificant interest in industry and academia as an alternative=0Aplatform for running scientific applications. Given the dynamic nature=0Aof IaaS clouds and the long runtime and resource utilization of such=0Aapplications, an efficient checkpoint-restart mechanism becomes=0Aparamount in this context. This paper proposes a solution to the=0Aaforementioned challenge that aims at minimizing the storage space and=0Aperformance overhead of checkpoint-restart. We introduce an approach=0Athat leverages virtual machine (VM) disk-image multi-snapshotting and=0Amulti-deployment inside checkpoint-restart protocols running at guest=0Alevel in order to efficiently capture and potentially roll back the=0Acomplete state of the application, including file system=0Amodifications. Experiments on the G5K testbed show substantial=0Aimprovement for MPI applications over existing approaches, both for=0Athe case when customized checkpointing is available at application=0Alevel and the case when it needs to be handled at process level. SUMMARY:BlobCR: Efficient Checkpoint-Restart for HPC Applications on IaaS Clouds using Virtual Disk Image Snapshots PRIORITY:3 END:VEVENT END:VCALENDAR