Partial join order optimization in the paraccel analytic database

The ParAccel Analytic DatabaseTM is a fast shared-nothing parallel relational database system with a columnar orientation, adaptive compression, memory-centric design, and an enhanced query optimizer. This modern object-oriented optimizer and its optimizer framework, known as Volt, provide efficient bulk and instance level query expression representation, multiple expression managers, and rule and cost-based expression transformation organized via multiple optimizer instances. Volt has been applied to the problem of ordering very large numbers of joins by partially ordering them for subsequent optimization using standard dynamic programming. Performance analyses show the framework's utility and the optimizer's effectiveness. Categories and Subject Descriptors H.2.4 [Database Management]: Systems ? parallel databases, query processing, relational databases. General Terms Algorithms, Performance, Design, Experimentation. Keywords Query Optimization, Extensible Optimizer Framew...

• Type: Conference
• Year: 2009
• Where: SIGMOD
• Authors: Yijou Chen, Richard L. Cole, William J. McKenna, Sergei Perfilov, Aman Sinha, Eugene Szedenits Jr.
• Keywords: Database, Enhanced Query Optimizer, Extensible Optimizer Framework, Multiple Optimizer Instances, SIGMOD 2009