A switching search proxy

Search engine has become a very effective tool to find wanted information in the web. Huge numbers of query requests are generated everyday. It is impossible to handle all these voluminous query requests in a single machine. Such problem is handled in general by duplicating the search service in many servers and putting a layer-4 switching device as the entry point of the duplicated query processing servers. The workload of query requests is therefore distributed to the duplicated query processing servers. However, a layer-4 switching device dispatches the query requests in a blind and random manner.In this paper, we design a new approach called Switching Search Proxy to replace the layer-4 switching device. Our switching search proxy uses a content sensitive query partitioning scheme to dispatch the query requests. Such scheme enables a very effective query caching mechanism composed of a dynamic caching and static caching to greatly improve the performance of the search service. In our experiments, we found that up to 55% improvement in response time can be achieved by our switching search proxy, compared to the traditional layer-4 switch approach.