Browse Prior Art Database

Parallel Data Partitioning of Shared Memory Computers, Using Fetch and Add With Block Transfers

IP.com Disclosure Number: IPCOM000057698D
Original Publication Date: 1988-Jun-01
Included in the Prior Art Database: 2005-Feb-15

Publishing Venue

IBM

Related People

Authors:
Franaszek, PA Heidelberger, P [+details]

Abstract

A technique is described whereby multiple processors can, in parallel, partition a data array that is stored in shared memory, into multiple data arrays that are also stored in shared memory. Fetch and add concepts are utilized in a parallel partitioning algorithm, so as to avoid processor synchronization delays. Also, block data transfers are used to minimize interconnection network delays. Typically, parallelism is used in data partitioning to improve performance of database computer systems.