CODISC: Collaborative and distributed semantic caching for maximizing cache effectiveness in wireless networks

In wireless mobile ad hoc networks (MANETs), a mobile node would normally acquire data from a data server through an access point by sending the server a request each time it needs data. To reduce the high costs normally associated with accessing remote servers (i.e., outside the MANET), data caching by the mobile nodes can be employed. Several caching techniques for MANETs have been proposed and implemented, including a cooperative scheme that we recently introduced. It employs a directory-based approach in which submitted queries are cached in the MANET to be used subsequently as indexes to corresponding data items (results). When a request is issued, nodes cooperate to find its answer (if it exists) and send it to the requesting node. In this paper, we extend this scheme by semantically comparing each submitted request with all cached queries. The semantic analysis process includes trimming the request into fragments and joining the answers of these fragments to produce the answer of the request. We study the performance of the proposed system both analytically and experimentally, and prove the advantageous features of the system relative to others in terms of query response time, generated traffic, and hit ratio.

Research highlights
► Queries are fragmented into subqueries that may be answered by caching nodes.
► In CODISC, a cache hit can occur even if the data is not found as one piece.
► CODISC maximizes cache utilization, and reduces the average query delay.
► A node that requested a query can obtain and start processing partial results.
► The average node energy consumption in the wireless mobile network is estimated.