Pipeline Stage Based Dynamic Load Balancing for Right-Deep Multi-Joins

スポンサーリンク

概要

• 論文の詳細を見る

• With ever increasing database sizes, parallel processing has become an important technology for improving over all query response time. However, this reduction in response time comes at the expense of the increased complexity required for distributing the work to be done in a query evenly amongst the processors. This paper presents our dynamic load balancing algorithm for right-deep hash multi-joins executing on a shared nothing architecture, which makes use of a centralized processor to handle all load balancing decisions. The overall goal is to assure that each processor generates nearly the same number of multi-join result tuples, and this is achieved through the dynamic migration of build relation tuples between processors.

• 一般社団法人情報処理学会の論文
• 1996-03-06

著者

• Davis Stephen

  University Of Tokyo Institute Of Industrial Science

• Kitsuregawa Masaru

  University Of Tokyo

• Kitsuregawa Masaru

  University Of Tokyo Institute Of Industrial Science

関連論文

• Relational Algebra Machine GRACE
• Join Strategies on Grid-Files
• Join Strategies on Multi-Dimensional C1ustered Relations
• On Parallel Hash-Join Processing with Skewed Data
• Pipeline Stage Based Dynamic Load Balancing for Right-Deep Multi-Joins

スポンサーリンク