Optimizing joins in fragmented database systems on a broadcast local network

The problem of optimizing joins between two fragmented relations on a broadcast local network is analyzed. Data redundancy is considered. Semantic information associated with fragments are used to eliminate necessary processing. More than one physical copies of a fragment is allowed to be used in a strategy to achieve more parallelism. Join-analysis graphs are introduced to represent joins on two fragmented relations. The problem of optimizing a join is mapped into an equivalent problem of finding a minimum-weight vertex cover for the corresponding join-analysis graph. This problem is proved to be NP-hard. A four-phase approach for processing joins is proposed