HTTP ACCELERATION BY PREDICTION AND PRE-FETCHING

Described is a technology by which high latency problems with respect to web requests are reduced by having a web proxy server predict and pre-fetch content, in parallel, that is to be requested by a client. The web proxy analyzes a main web page requested by a client to predict subsequent client requests. The web proxy pre-fetches content before the client requests it, by making concurrent requests for the page's embedded objects that exceed the client's limited number of (e.g., two) connections. In one example, the web proxy sends HTTP requests substantially in parallel to a web server, thereby reducing overall latency. In another example, the web proxy server sends parallel requests to a remote web proxy coupled to a web server. The remote web proxy requests only a limited number of objects (e.g., two) at a time, but does so over fast (low latency) connections to the web server.