Pre-fetch Engines

The local pre-fetch engine's main responsibility is to issue pre-fetch requests to the local cache manager based on the lists obtained from the local and backbone profile engines. The recommendation list that arrives from the backbone profile manager goes through the local filter first, in case there are some documents that the individual wishes to block out. It then amortizes the pre-fetch requests over time so that the local cache manager does not get overloaded. At the same time, the local pre-fetch engine makes sure that duplicate items are not pre-fetched.

A special case of pre-fetch is hoarding, which is done only by the local pre-fetch engine. In the case of hoarding, only the node weights are used in order to pre-fetch the documents that the user is most likely to require in the future. As in the case of disconnected file systems, this gives rise to the problem of caching versus hoarding. An approach similar to hoard-walking [14] is used in order to refresh commonly used items.

It should be noted that on the local machine, user requests are given priority over pre-fetch requests. That is, no pre-fetch requests are issued until there are no pending user requests.

The backbone pre-fetch engine's main responsibility is to issue pre-fetch requests based on the list obtained from the backbone profile engine. Note that multiple lists can be submitted to the backbone pre-fetcher if there are multiple users logged on. The backbone pre-fetcher will ensure that duplicate items are not pre-fetched and that pre-fetch requests are amortized over time.