Network　measurement　has　received　a　lot　of　attention　during　the　past　few　years　because　the　rapid　growth　of　the　Internet　has　caused　many　internetworking　problems　in　network　administration,　routing　and　resource　management.　For　network　measurement,　the　placement　of　measurement　nodes　is　the　key　and　the　study　of　placement　algorithms　often　requires　analysis　and　simulation　using　a　network　model　that　resembles　real　networks.　In　this　paper,　we　study　the　problem　of　choosing　the　network　measurement　nodes　in　which　we　turn　the　problem　into　a　topology-clustering　graph　problem　and　propose　a　solution　based　on　a　nearest　dynamic　clustering　algorithm.　Our　algorithm　starts　with　the　initial　cluster　centers　whose　number　is　normally　several　times　that　of　the　requisite　measurement　nodes.　Then,　we　cluster　the　network　nodes　by　the　minimum　link　distance　computing,　and　the　nodes　that　have　the　nearest　link　distance　to　each　node　in　every　cluster　are　selected　as　the　new　cluster　centers.　Last,　we　incorporate　the　near　clusters　by　the　link　distance　of　each　cluster　centers　and　dynamically　change　the　cluster　centers.　The　process　continues　as　long　as　the　new　clusters　can　support　the　measurement　requirements.　When　the　process　completes,　the　final　cluster　centers　are　chosen　as　the　measurement　nodes.　Simulation　results　show　that　our　clustering　algorithm　can　satisfy　the　three　evaluation　metrics:　Cluster　variance,　cluster　centers　distance　and　average　cluster　radius.　©　2008　Binary　Information　Press.