Techniques and data structures for parallel resource management
This thesis contributes to parallel resource management in two ways. First considered is management of system state for computation structures consisting of arrays of computations that differed only in indexing parameters. Simple decompositions of the externally visible state into simultaneously updatable components are proposed. Second considered is management of system state for weakened priority queues. The two priority structures proposed, a concurrent heap and a software banyan, were found to be efficient and effective. In addition, the author contributed in the area of language tools for computations that utilize predefined abstract-data-type implementations. A mechanism for abstract data type definition is presented. To promote simultaneity of update, a significant extension of the linguistic construct of path expressions is defined and used as a basis for defining implementation of sequencing within abstract data types. The main advantage of using extended path expressions is that, in addition to synchronization requirements, binding of activities to object decompositions may be specified, along with runtime consistency checking, while leaving the object implementation to the underlying system.
- Research Organization:
- Texas Univ., Austin (USA)
- OSTI ID:
- 6355037
- Resource Relation:
- Other Information: Thesis (Ph. D.)
- Country of Publication:
- United States
- Language:
- English
Similar Records
A Fault Oblivious Extreme-Scale Execution Environment
A Fault-oblivious Extreme-scale Execution Environment
Related Subjects
DATA BASE MANAGEMENT
PARALLEL PROCESSING
ABSTRACTS
ALGORITHMS
DATA PROCESSING
MODIFICATIONS
PROGRAMMING LANGUAGES
SYNCHRONIZATION
DOCUMENT TYPES
MANAGEMENT
MATHEMATICAL LOGIC
PROCESSING
PROGRAMMING
990210* - Supercomputers- (1987-1989)
990300 - Information Handling