Cluster-based web servers are a popular approach to achieving high performance for the ever-growing web traffic. Most existing systems use a Layer 4 dispatcher to assign client requests to the web server nodes. Some content-based dispatching solutions need to parse the HTTP request in order to make the dispatching decision. They either can achieve only coarse-grain load balancing or would introduce heavy load on the dispatcher. Cooperative caching allows the requested document to be fetched from a peer server node\222s in-memory cache, rather than from the local hard disk. While this can greatly improve the performance, it consumes much network resources when transferring large cache objects. In this paper, we present a new network support mechanism, called Socket Cloning (SC), in which an opened socket can be migrated efficiently between cluster nodes. With SC, the processing of HTTP requests can be moved to the node that has a cached copy of the requested document, thus bypassing any object transfer between peer servers. A prototype has been implemented and micro-benchmark tests show that SC incurs much less overhead than all the mentioned approaches for any file size. In trace-driven benchmark tests, our system can outperform other approaches by 90%.