학술논문
GLU: a system for scalable and resilient large-grain parallel processing
Document Type
Conference
Author
Source
Proceedings of the Twenty-Fourth Annual Hawaii International Conference on System Sciences System Sciences, 1991. Proceedings of the Twenty-Fourth Annual Hawaii International Conference on. i:36-48 vol.1 1991
Subject
Language
Abstract
Describes a system called GLU for large-grain parallel processing. GLU provides for ease of programming, scalable performance, and resilient execution. GLU is a hybrid system that combines the powerful abstract concepts of declarative systems with the practical concrete aspects of procedural systems. In the GLU programming model, a program is declarative at the higher levels and procedural at lower levels. The GLU abstract architecture for executing GLU programs is based on the demand-driven dataflow execution model. This execution model facilitates automatic exploitation of large-grain parallelism inherent in a program and provides resilience from partial hardware failures.ETX