This paper analyzed the features of a new network architecture whose locator and identifier are separated, in order to improve the query and replacement efficiency of the identifier mapping, the backup of the locator/identifier mapping information needs storing in the most necessary place, which is a problem of cache management in fact .In this paper the typical cache management algorithms are analyzed firstly, and then a new aimed pushing and pre-fetching strategy is proposed according to the bidirectional interactive character of the communication activity in querying the mapping relationship, thus divided the Access Switch Router(ASR) cachespace into three section: Waiting_First_Access_Section (WFA), Frequently_Used_se- ction(FU) and Session_Duration_Section (SDU). And based on the identity mapping information reuse probability predicted by the Markov model, an information lifetime-based adaptive cache management algorithm is proposed. The simulation result shows that the performance of the buffer management algorithm proposed by this paper is better than those of the existing cache management algorithms such as LFU, LRU and LFU-LRU
Adaptive Cache Management for Energy-Efficient GPU Computing