Abstract: A method comprises receiving a first request for a first segment of a content stream in a network element from a given one of a plurality of clients, determining in the network element whether the first segment is stored in a memory of the network element, sending a second request for the first segment from the network element to a server responsive to the determining step, receiving a response comprising the first segment in the network element from the server responsive to the second request, and sending the first segment from the network element to the given one of the plurality of clients. The first segment is related to a second segment of the content stream, the relationship being transparent to the network element but being inferable based at least in part on at least one of the first request, the response and one or more prior requests.