- Abstract:
-
Recently there has been a growing interest in the research community on join query evaluation for scenarios in which inputs enter the system at highly variable and unpredictable rates. The focus in such cases shifts from completing the computation as soon as possible to being able to produce a prefix of the output as soon as possible. To tackle the problem, we propose a multi-input symmetric join operator we call MJoin. Employing a single operator is in contrast to the proposed solutions to date, which rely upon some combination of streaming binary operators and "on-the-fly" query plan reorganization to deal with rates that are unpredictable. We built a prototype implementation of the MJoin operator and experimented with traditional joins, considering workloads consisting of inputs having an arrival rate with a steady mean value as well as with a fluctuating mean value. Our results show that in many instances the MJoin produces outputs sooner than any tree of binary operators. Additionally, execution plans consisting of MJoins are fixed, thus alleviating the need for frequent plan adaptation. This suggests that the MJoin operator may be a useful addition to systems in which the input arrives as streams from remote sites.
- Links To Paper
- No links available
- Bibtex format
- @InProceedings{EDI-INF-RR-0597,
- author = {
Stratis Viglas
and Jeffrey Naughton
and Josef Burger
},
- title = {Maximizing the Ouput Rate of Multi-Way Join Queries over Streaming Information Sources},
- book title = {Proceedings of VLDB 2003 (Very Large Data Bases Conference)},
- publisher = {Morgan Kaufmann},
- year = 2003,
- month = {Sep},
- pages = {285-296},
- }
|