Abstract: In-net caching is an important part of the content centric network architecture. So far, the default "Leave Copy Everywhere (LCE)" caching policy will cache all content without distinction, lead to a large number of cache content redundant in the nodes of network. Due to the limitation of node cache size, redundant reduced the verity of cache content cached in network, result in bad utilization of cache space. To boost the performance of the cache, many improved mechanisms have been put forward, including LCD, MCD and ProbCache caching policies. This article respectively experimented the LCE, LCD, MCD and ProbCache algorithm, judging by standards like content response delay, network link load, server stress, evaluation index, analyzed factors, such as network topology, content popularity and cache space, have what kind of effect on these policies’ performance.