Heurísticas para balanceamento de carga de máquinas em infraestruturas de nuvem.

Abstract

In CloudComputingenvironments,especiallythoseusingtheinfrastructureasaservice model, theelasticitycharacteristicinresourceprovisioningcomeswiththeneedtomanage resources sothequalityofservicecancontinuetobeguaranteedtousersandalsoto maintain agoodperformanceoftheinfrastructure.Thisworkproposesheuristicsthat are abletoassistintheloadbalancingoftheserversinaCloudinfrastructure,proposing migrations toreducetheoverheadintheserversthatwereidentifiedasoverloaded,since with thepassageoftimethereisanaturalvariationintheamountofresourcesinuse.This variationinaconsequenceofremovaloradditionofapplicationsandevenoftheusageof verticalscalingtoimproveapplication’sperformance.Atoolwasimplementedtomake use oftheheuristicalgorithmsandthustoaidintheexperimentsandtheirvalidation,the metrics usedcomedirectlyfromheterogeneousserversoftheOpenStackCloudofthe DistributedSystemsLaboratory.TheresultsshowthatinadditiontothedecreaseinCPU consumption ofserversthatwereoverloaded,itisalsopossibletoimprovetheperformance of theseserversinsomecases.