Query Load Balancing in Parallel Database Systems

Luc Bouganim 1
1 PETRUS - Personal Trusted cloud
Inria Saclay - Ile de France
Abstract : Definition : The goal of parallel query execution is minimizing query response time using inter- and intraoperator parallelism. Interoperator parallelism assigns different operators of a query execution plan to distinct (sets of) processors, while intraoperator parallelism uses several processors for the execution of a single operator, thanks to data partitioning. Conceptually, parallelizing a query amounts to divide the query work in small pieces or tasks assigned to different processors. The response time of a set of parallel tasks being that of the longest one, the main difficulty is to produce and execute these tasks such that the query load is evenly balanced within the processors. This is made more complex by the existence of dependencies between tasks (e.g., pipeline parallelism) and synchronizations points. Query load balancing relates to static and/or dynamic techniques and algorithms to balance the query load within the processors so that the response time is minimized.
Document type :
Book sections
Complete list of metadatas

Cited literature [12 references]  Display  Hide  Download

https://hal.inria.fr/hal-01660649
Contributor : Luc Bouganim <>
Submitted on : Monday, December 11, 2017 - 11:17:13 AM
Last modification on : Monday, January 15, 2018 - 8:01:03 AM

File

B1.pdf
Files produced by the author(s)

Identifiers

Citation

Luc Bouganim. Query Load Balancing in Parallel Database Systems. L. Liu; M.T. Özsu. Encyclopedia of Database Systems (2nd edition), Springer, pp.1-6, 2017, ⟨10.1007/978-1-4899-7993-3_1080-2⟩. ⟨hal-01660649⟩

Share

Metrics

Record views

117

Files downloads

255