A novel web caching scheme using hybrid least frequently used and support vector machine

Abstract

The gargantuan uses of web access in various types of applications, such as text, image, audio, and video, across the globe have caused the limitation for service providers to optimally make use of Internet infrastructure. The advance of web proxy/caching has recently been in place to mitigate this phenomenon using the concept of locality and proximity. There exist some traditional caching schemes, such as FIFO, LFU, LRU, and Size, but with key limitations on the precision. On the other hands, soft computing has recently been investigated due to its advantage of high precision. Thus, this paper proposes a novel caching method by integrating these twos. SVM was first used for classification, to divide the caching probability — to be replaced or else. Then, LFU was applied for the actual replacement given new web objects (if cache full); and these are Hybrid LFU-SVM. Its performance is practically confirmed from our intensive evaluation against SVM-LRU and its traditional schemes like LFU and LRU in order of 14% to 52.3% and 18% to 63.2%, for hit and byte hit rate, respectively, using a standard NLANR dataset.