next previous
Next: Criteria for Comparison Up: Scalable Distributed Dynamic Grids: A Previous: Scalable Distributed Dynamic Grids: A

Introduction

Computational grids are used by many numerical techniques as a means of discretizing the continuum domain. The parallelization of these techniques typically consists of partitioning the computational grid across processing elements. Managing creation, deletion, accesses, and updates for such a distributed grid (or array) structure on existing distributed memory muti-processor systems is non-trivial and requires low-level communications, synchronization, and resource management; the details of which need to be hidden from the application developer. To this end, a number of software system have been developed to provide high-level support for defining and using distributed grid or array data-structures. The objective of this report is to survey these software systems and to compare their functionality.

Section 2 describes system features compared in this survey and is followed by a description of related systems.



Manish Parashar, parashar@cs.utexas.edu