[Zurück]


Beiträge in Tagungsbänden:

M. Schlaipfer, K. Rajan, A. Lal, M. Samak:
"Optimizing Big-Data Queries Using Program Synthesis";
in: "Symposium on Operating Systems Principles", ACM, 2017, ISBN: 978-1-4503-5085-3, S. 631 - 646.



Kurzfassung englisch:
Classical query optimization relies on a predefined set of rewrite rules to re-order and substitute SQL operators at a logical level. This paper proposes Blitz, a system that can synthesize efficient query-specific operators using automated program reasoning. Blitz uses static analysis to identify sub-queries as potential targets for optimization. For each sub-query, it constructs a template that defines a large space of possible operator implementations, all restricted to have linear time and space complexity. Blitz then employs program synthesis to instantiate the template and obtain a data-parallel operator implementation that is functionally equivalent to the original sub-query up to a bound on the input size.

Program synthesis is an undecidable problem in general and often difficult to scale, even for bounded inputs. Blitz therefore uses a series of analyses to judiciously use program synthesis and incrementally construct complex operators.

We integrated Blitz with existing big-data query languages by embedding the synthesized operators back into the query as User Defined Operators. We evaluated Blitz on several production queries from Microsoft running on two state-of-the-art query engines: SparkSQL as well as Scope, the big-data engine of Microsoft. Blitz produces correct optimizations despite the synthesis being bounded. The resulting queries have much more succinct query plans and demonstrate significant performance improvements on both big-data systems (1.3x --- 4.7x).


"Offizielle" elektronische Version der Publikation (entsprechend ihrem Digital Object Identifier - DOI)
http://dx.doi.org/10.1145/3132747.3132773


Erstellt aus der Publikationsdatenbank der Technischen Universität Wien.