In　recent　years,　as　an　emerging　technology,　cloud　computing　has　provided　us　with　convenient　services,　and　power　consumption　on　issues　have　become　increasingly　prominent.　Virtual　machine　live　migration　technology　has　become　an　important　technology　to　reduce　the　power　consumption　of　cloud　computing　centers.　In　the　process　of　virtual　machine　migration,　the　performance　of　the　virtual　machine　is　inevitably　degraded,　which　may　violate　service　level　agreement　(SLA,　Service　Level　Agreement).　How　to　use　virtual　machine　live　migration　technology　to　reduce　power　consumption　as　much　as　possible　while　ensuring　a　low　SLA　violation　rate　becomes　a　hot　issue.　This　paper　aims　to　optimize　the　light　load　detection　and　virtual　machine　redistribution　in　the　virtual　machine　live　migration　model.　Aiming　at　the　problem　that　the　existing　virtual　machine　light　load　detection　method　is　easy　to　cause　＇over-migration＇,　this　paper　proposes　a　threshold-based　minimum　CPU　utilization　method　for　light　load　detection,　which　effectively　avoids　excessive　virtual　machine　migration.　Aiming　at　the　problem　that　the　current　process　of　virtual　machine　re　allocation　algorithm　is　relatively　simple,　and　there　is　a　certain　power　loss　space,　we　present　power　aware　simulation　annealing　algorithm　(PASA).　The　algorithm　combines　the　simulated　annealing　algorithm　based　on　the　power　aware　best　fit　decreasing　algorithm　(PABFD),　which　largely　avoids　the　disadvantage　that　the　PABFD　easily　falls　into　the　local　optimal　solution　trap.　The　paper　uses　the　CloudSim　simulator　as　simulation　platform.　The　results　show　that　compared　with　the　best　algorithm　combination　proposed　by　the　previous　researchers,　the　power　consumption　of　the　new　algorithm　combination　proposed　in　the　paper　is　reduced　by　16.79%,　and　the　SLA　violation　rate　is　reduced　by　85.37%.　Combining　the　two　algorithms　together　can　lead　to　better　energy　efficiency,　performance　and　quality　of　service　than　using　the　two　algorithms.　©　Springer　Nature　Singapore　Pte　Ltd.　2018.