Cooperative web caching using dynamic interest-tagged filtered bloom filters

Although cooperative Web caching has been widely researched, comparatively little has been done to reduce inter-proxy network overhead whilst allowing for a high percentage of requested documents to be retrieved from the cache. Alleviating these issues can substantially reduce web traffic, increase scalability and enhance a user's browsing experience. This paper introduces a novel cache sharing system employing data structures called Dynamic Interest-Tagged Filtered Bloom Filters (DITFBFs). DITFBFs are capable of representing the cache content of a proxy in a compact form, which is then shared with other proxies in the cooperative Web caching system. What distinguishes the proposed system from others is that DITFBFs only represent the portion of a proxy's cache content that will be of interest to another proxy. This then results in a reduction of inter-proxy overhead. Experimental simulations indicate that, when compared with existing protocols, the proposed system is capable of multiple improvements. Namely, lowering the number of remote cache search messages by at least 60%, decreasing user-perceived latency by at least 65% and appreciably reducing the overall inter-proxy network overhead. The proposed system accomplishes this whilst maintaining a cache hit ratio as high as the other protocols.