General Caching with Lifetimes

We consider the problem of caching with lifetimes, where a lifetime is specified whenever a page is loaded into the cache. The copy of a page loaded into the cache may be used to serve requests to the same page, only until its expiration time. We present a generic method to get an algorithm for caching with lifetimes, from an algorithm for caching without lifetimes. This method works for any cost model, and in online as well as offline settings. In the online (resp. offline) setting, the competitive (resp. approximation) ratio of resulting algorithm for caching with lifetimes, is one more than the competitive (resp. approximation) ratio of the original algorithm for caching without lifetimes. Using this method and the existing algorithms for caching without lifetimes, we get an Hk + 1 competitive randomized algorithm and a 2-approximation algorithm for standard caching with lifetimes, where k is the cache size. This is an improvement over the (2Hk + 3)-competitive algorithm and the 3-approximation algorithm given by Gopalan et. al. [GKM02]. We also get O(log k) competitive randomized algorithms for various subclasses of general caching such as weighted caching and the Bit and Fault models, asymptotically matching the lower bound of Hk on the competitive ratio; and a 5 approximation algorithm for the general offline caching problem.