Are you an EPFL student looking for a semester project?
Work with us on data science and visualisation projects, and deploy your project as an app on top of Graph Search.
With the Internet success leading to heavy demands on network, proxies have beco me an unavoidable necessity. In this paper we present a new technique to improve caching services for an Internet user group. We propose an alternative to classical LRU, LFU or similar algorithms. Our appro ach is based on object usage, cross-references and geographic location. With our system we will not only improve on storing performance taking into account pref erences and usage of specific user groups, but we will also improve latency perf ormance with accurate pre-fetching. In order to do this, we use three mechanisms: (1) a normalised indexing tree bas ed on the URL semantic; (2) a graph representing inter-documents references; (3) a Kohonen like algorithm to bring out the real topology of the web usage space.
Martin Jaggi, Thijs Vogels, Hadrien Hendrikx
Anne-Marie Kermarrec, Erick Lavoie