SKEW HANDLING STRATEGIES FOR PIPELINED PROCESSING OF MULTI-JOIN QUERIES IN SHARED-NOTHING SYSTEMS

This paper looks at how to effectively exploit pipelining for multi-jo in queries in shared-nothing systems. A promising technique that has a ppeared in the literature uses an iterative approach to process a mult i-join query. In each iteration, several relations are selected, and a re joined in a pipelined fashion. However, optimization algorithms tha t are based on this approach have traditionally assumed that the relat ions are uniformly distributed or lowly skewed. When this assumption i s relaxed, that is when the data is skewed, the performance of the sys tem may degenerate drastically. We propose four skew handling techniqu es to deal with data skew for multi-join queries. An analytical model is presented and used to compare the performance of each technique. Ou r results show that the nested-loops technique performs worst, while t he hybrid technique is superior in most cases.