Development of a virtualized supercomputing environment for genomic analysis

Abstract

Recentlytheimportanceofgenomicdataanalysishasbeengrowing,onerealizesnecessityofthepersonalizedtreatmentofhumancancers.Nextgenerationsequencing(NGS)techniqueisacost-effectivewaytoobtainsuchdatasetsforcancerdataanalysis.Hence,mostofbioinformaticsresearchgroupsusetheNGStechniquetoobtainsuchdatasets.TheamountofNGSdataishugeandrapidlygrowing,therefore,itrequiressupercomputingsystemstobehandledwithinareasonabletime.BioinformaticsresearchersanalyzethesetsbyusingNGSapplicationssuchasBWAandBowTie,butthoselegacyapplicationshavelimitedscalabilityandresourceutilizationonsupercomputingsystems.

Toresolvethissituation,wedevelopedavirtualizedtechniquebyimprovingtheresourceutilizationandscalabilityofNGSapplications.First,toimproveresourceutilization,thevirtualizedsystemarchitectureisbuiltbyallocatingvirtualmachinesconsideringthelimitationofresourceutilization.Second,thevirtualizedsystemarchitectureconsideringdatalocalityispresentedtoimprovescalability.Finally,experimentalresultsshowthatourvirtualizedsystemachievedapproximately30%betterperformancethannativesystems.Inaddition,theperformanceofthesystemconsideringdatalocalityachievesaspeeduptwicethatofasystemusingasingle-storageserver.