A cache server for providing content includes a processor configured to receive a first datagram from a client system sent to an anycast address, send a response datagram to the client system in response to the first datagram, receive a request datagram from the client system sent to the anycast address, and send a batch of content datagrams to the client system. The first datagram includes a universal resource locator corresponding to the content. The response datagram includes a content identifier for the content. The request datagram includes the content identifier, an offset, and a bandwidth indicator. The batch of content datagrams includes a portion of the content starting at the offset.