A load balanced multicomputer relational database system for highly skewed data
| Autor: | Ophir Frieder, Fotis Barlos |
|---|---|
| Rok vydání: | 1995 |
| Předmět: |
Computer Networks and Communications
Computer science Relational database Distributed computing ComputingMethodologies_IMAGEPROCESSINGANDCOMPUTERVISION Parallel computing Query optimization computer.software_genre Computer Graphics and Computer-Aided Design Partition (database) Theoretical Computer Science Relational database management system Artificial Intelligence Hardware and Architecture Scalability Sargable Hypercube computer Software |
| Zdroj: | Parallel Computing. 21:1451-1483 |
| ISSN: | 0167-8191 |
| DOI: | 10.1016/0167-8191(95)00028-m |
| Popis: | Query optimizers are used in most modern high performance databases. The emergence of multiprocessor machines allows the development of more aggressive database designs, which in turn need very complicated query optimizers. We present the Workload Partitioning Segment of DOME, a query optimization environment developed on an Intel i860 hypercube system. DOME calculates the distribution of the query input relations and uniformly allocates the workload associated with the resolution of a query on the nodes of a multiprocessor system. The workload is uniformly allocated even when the input relations exhibit a high degree of skew. A priori information of the degree of skew is unnecessary. DOME provides routines to perform the relational operations in parallel on all nodes of the hypercube system. We tested DOME with various sizes, and degree of skew of the input relations. We describe the theoretical foundation behind the optimization techniques employed and present graphs that portray the performance of DOME and its scalability. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|
Full Text from ScienceDirect