A PRE-FETCH MECHANISM FOR PROXY SERVERS: USING ASSOCIATION RULES

Abstract

We consider the problem of decreasing users' waiting time in accessig WWW objects. The traditional soulutin uses the proxy server to cache remote objects. In this study, we propose a pre-fetch mechanism which uses the technique of mining association rules to improve the cache hit ratio of the Squid proxy server. According to the requests made by proxy server users , this mechanism can pre-fetch WWW objects that are probably accessed next and store them in the local cache of the proxy server. The user can then fetch the next objects in a short time. To verify the feasibility of this pre-fetch mechanism, we use user accesslogs in the proxy server to test the performance of our mechanism. Experimental results show an improvement of about 30% of the cache hit ratio, and learning more association rules can even increase the cache hit ratio about 47%.