Multi-phase interval caching for a news on demand server

We propose a buffer cache algorithm to build a cost effective storage server for real time news objects. In a news on demand server the number of concurrent streams is bounded for the maximum disk bandwidth the NOD server provides. The maximum numbers of concurrent streams can be increased more than the total disk bandwidth supports if the buffer cache is applied to hold hot news objects. However it is said that conventional caching algorithms do not go well with real time news objects. With our proposed caching scheme all of the news objects are classified based on their arrival rate; their eligibility is then checked for caching. Only suitable objects are permitted to access the buffer pool to reduce unnecessary buffer replacement. We use simulation to verify the efficiency of our proposed algorithm and our results show that by using this caching scheme, compared with servicing requested streams only by disks, a greater number of requests are serviced without extra cost.