With　the　development　of　the　cloud　computing,　more　and　more　devices　will　be　connected　to　the　network,　and　bring　enormous　load　to　current　network.　Currently,　the　data　of　devices　are　processed　in　the　cloud,　which　are　normally　located　far　away　from　the　devices.　Therefore,　network　bandwidth　and　communication　latency　become　serious　bottlenecks.　Edge　computing,　with　its　advantage　that　offers　cloud-like　services　at　the　edge　of　network　and　less　latency,　has　become　a　new　direction　of　the　current　Internet　of　Things　network　transmission　architecture.　How　to　schedule　the　tasks　between　the　edge　servers　and　the　central　cloud　server　so　that　the　task　response　time　can　be　minimized　becomes　the　critical　problem　in　edge-cloud　system.　In　this　paper,　we　propose　a　general　algorithm　to　resolve　this　problem.　In　our　model,　the　edge　servers　based　on　the　aware　of　tasks　expected　completion　time　to　decide　sending　the　tasks　to　the　other　edge　servers　or　cloud　when　it　is　fully　loaded　to　make　the　most　use　of　the　edge　servers　in　the　same　area.　We　evaluate　our　proposed　policy　in　iFogSim　toolkit.　Results　of　the　simulation　demonstrate　that　our　strategy　improves　significantly　in　reducing　the　latency　of　application　and　network　usage.　©　2019　IOP　Publishing　Ltd.　All　rights　reserved.