近年有些index caching的機制被應用於改善Gnutella-like P2P網路的scalability,其中部份research也修改了網路架構以及搜尋(search)方式以配合其index caching的機制。Index caching的方式可以大幅減少搜尋受歡迎物件的traffic,但是在網路上存有過多相同的index是其一個問題點。 在我們的Active Index Diffusing scheme,我們嘗試將各物件的index數量作平均化。被要求率接近的資源,他們的index會被群聚在一起,搜尋也會優先連去存有較高被要求率的index的節點。模擬的結果顯示出我們的架構可以提供更高的搜尋成功率比,搜尋的花費也是有顯著的減少。
In order to improve the scalability and reduce the traffic of Gnutella-like unstructured peer-to-peer networks, index caching mechanisms had been discussed in recent years. Some of them had modified the network architecture and adapted existing search methods for their mechanisms. Although the traffic of querying popular keywords could be reduced by using index caching mechanisms, the cached indices too redundant replications in the whole network. We try to make the amount of indices fair for each file name in our index diffusing scheme. In our scheme, the indices with similar request rate would be grouped together, and the peers of the group who have indices with higher request rate will receive query messages earlier. In the simulations, our architecture could perform queries with higher success rates than existing research in multilayer peer-to-peer networks, and the traffic is also decreasing conspicuously.