HitBand: A Prefetching Model to Increase Hit Rate and Reduce Bandwidth Consumption

Caching is a very important issue in distributed web system in order to reduce access latency and server load. A request is a hit if it is available in the cache and if not then it will fetch from the server in order to cache and serve the request. Researches have shown that generic algorithms of caching can increase hit rate up to 40−50%, but adding prefetching scheme can increase this rate to 20%. Prefetching is a technique to fetch documents before they are requested. This paper proposes a process model for prefetching named HitBand which will balance hit rate bandwidth in every scenario with the combination of "Roulette-wheel selection". Roulette-wheel selection is a very popular selection based algorithm which selects objects according to their fitness. We have compared our HitBand with the generic algorithms of prefetching like prefetching by popularity, apl characteristic, good Fetch and lifetime. Generic algorithms did not take web object size into consideration and in limited bandwidth scenario object size has a big impact on bandwidth consumption. Though prefetching by lifetime algorithm shows little concern about bandwidth consumption by getting the object with changes happening less frequently but this compromises the hit rate. But our proposed HitBand not only considers bandwidth but also hit rate during prefetching. Performance evaluation of HitBand along with other algorithms is provided in our paper. We have tested our HitBand with the testing engine which is built using JavaScript and maintained under AngularJS framework. From the performance evaluation, our HitBand shows better results both in high and low bandwidth.