To　tackle　the　problem　that　the　internet　can　not　adapt　to　the　requirements　of　ever-changing　and　wide　applications,　a　novel　evolutionary　network　model　is　proposed.　The　evolutionary　network　can　adapt　its　behaviors　according　to　the　network　conditions　and　status　by　adaptively　changing　its　control　parameters　and　control　algorithms　to　achieve　different　QoS　for　different　applications.　Three　levels　of　evolution　including　parameter　adaptation,　network　module　reconstruction,　and　collaboration　among　network　nodes　are　presented.　The　architecture　of　the　evolutionary　networks　is　designed　and　a　software-based　prototype　is　developed.　Key　technologies　and　optimization　methods　for　performance　are　studied.　Experiment　results　show　that　the　evolutionary　network　can　adapt　to　the　dynamic　changes　in　network　conditions　and　provide　stable　and　efficient　network　service.