|Title||:||Holding the Query Optimizer SpillBound|
|Speaker||:||Jayant Haritsa (Indian Institute of Science, Bangalore)|
|Details||:||Fri, 2 Dec, 2016 3:00 PM @ BSB 361|
|Abstract:||:||To address the classical selectivity estimation problem in databases, a radically different approach called PlanBouquet was recently proposed, wherein the estimation process is completely abandoned and replaced with a calibrated discovery mechanism. The beneficial outcome of this new construction is that, for the first time, provable guarantees are obtained on worst-case performance, thereby facilitating robust query processing.
The PlanBouquet formulation suffers, however, from a systemic drawback - the performance bound is a function of not only the query, but also the optimizers behavioral profile over the underlying database platform. As a result, there are adverse consequences: (i) the bound value becomes highly variable, depending on the specifics of the current operating environment, and (ii) it becomes infeasible to compute the value without substantial investments in preprocessing overheads.
In this talk we present SpillBound, a new query processing algorithm that retains the core strength of the PlanBouquet discovery process, but reduces the bound dependency to only the query. Specifically, SpillBound delivers a worst-case multiplicative bound of D2 + 3D, where D is simply the number of error-prone predicates in the user query. Consequently, the bound value becomes independent of the optimizer and the database platform, and the guarantee can be issued just by inspecting the query, without incurring any additional computational effort.
We go on to prove that SpillBound is within an O(D) factor of the best possible deterministic selectivity discovery algorithm in its class. Further, a detailed empirical evaluation over the standard TPC-H and TPC-DS benchmarks indicates that SpillBound provides markedly superior worst-case performance as compared to PlanBouquet in practice. Therefore, in an overall sense, SpillBound offers a substantive step forward in the quest for robust query processing.
[Joint work with Srinivas Karthik (IISc), Sreyash Kenkre and Vinayaka Pandit (IBM IRL).]
Speaker Bio: Jayant Haritsa is with the Department of Computer Science & Automation at the Indian Institute of Science, Bangalore, where he has been a professor of database systems since 1993. He received a B Tech degree from IIT Madras, and the MS and PhD degrees from the Univ. of Wisconsin-Madison. He is a Fellow of ACM and IEEE.