Efficient　use　of　the　distributed　resources　(computers,　databases,　and　human　expertise,　etc.)　requires　that　the　computational　load　must　be　balanced　across　processors　in　a　way　that　minimizes　communications　among　processors　and　data　moving.　Multilevel　partitioners　such　as　Minex　have　been　used　for　grid　computing　application　in　recent　research　works.　In　this　paper,　we　present　a　novel　map　algorithm　called　HACP,　an　ACO　algorithm　coupled　with　a　local　search.　ACO　algorithm,　due　to　their　intrinsically　distributed　and　multi-agent　nature　that　well　matches　these　types　of　architectures,　can　be　very　effective.　To　efficiently　solve　computational　load-balance　problems　we　give　a　parallel　implement　model　for　HACP.　Experimental　results　demonstrate　that　RACP　is　much　more　efficient　than　the　partitioning　method　in　MinEx.