A content delivery method and system, where the content delivery method includes receiving, by a first virtual server, a first request message from a content requester, where the first request message carries a first uniform resource locator (URL) of user-requested content, determining, by the first virtual server according to a cache list maintained by the first virtual server, that one of one or more storage servers managed by the first virtual server has cached the content corresponding to the first URL, and redirecting the first request message to a first storage server that is recorded in the cache list and that has cached the content corresponding to the first URL. A computation capability and a storage capability of a server cluster are separated, and content input/output (I/O) throughput performance is effectively improved.