Abstract: Wide-area distribution raises significant performance problems for traditional query processing technictues as data access becomes less predictable due to link congestion, load imbalances, and temporary outages. Norrblocking joining query execution is a promising approach to coping with unpredictability in unreliable network and hiding intermittent delays in data arrival by reactively scheduling background processing. Classical non-blocking two-way joining techniquc such as XJoin fail to deliver acceptable performance in such a scenario where gradually augmenting partition could not be dealt with during one relatively short intermittent delay. We developed a novel reactively-scheduled non-blocking join, called XJoin-FG，disparting one coarse-grained transaction into several parts according to the size of interval time.XJoin-FG employed fincgrained timestamp mechanism to avoid duplicate results. Using the optimization implementation along with emulational data obtained by monitoring Internet data delivery,we show that XJoin-FG is an effective solulion for providing fast ctuery responses to users even in the presence of the longer-term of data sources appeared as unavailability.

Key words: Join,Fine-grained,Non-blocking,Unreliable network