Low　transmission　speed　between　nodes　as　well　as　the　high　data　transferring　cost　are　ones　of　factors　leading　to　the,bottleneck　in　a　distributed　data　stream　system.　A　load　balance　algorithm　respecting　network　state　is　presented.　By　executing　corresponding　load　transferring　strategy　for　nodes　in　overload　or　under-load　state　respectively,　the　system　performance　can　be　greatly　improved.　The　algorithm　is　described,　and　the　test　result　is　given　and　analyzed.