Reliable task allocation in a fault-tolerant distributed system
The problem addressed in this thesis is that of assigning processes of a job among processors in a distributed computing system. A new classification and notation of the problem are introduced. The load balance among processors and the allocation reliability are addressed as key issues in the task allocation model. Other closed problems modeled are minimum reliability and maximum reliability problems. Algorithms are devised to compute the minimum reliability and maximum reliability allocation. The Kohler and Steiglitz branch-and-bound algorithm is modified for the task allocation model as an optimal solution. Since the task allocation problem is NP-complete, constraints, bound cost functions, and attribute dependent initial solutions are designed to eliminate the exhaustive search. Multiple heuristic solutions are developed. The average heuristic solution is based on three phases: get the initial allocation based on the average values of the process and processor attributes; reassign some of the processes to make the initial solution to be feasible if it is not; and allow redundancy to improve the allocation cost and the allocation reliability.
- Research Organization:
- Illinois Inst. of Tech., Chicago (USA)
- OSTI ID:
- 6721284
- Country of Publication:
- United States
- Language:
- English
Similar Records
Optimal resource allocation under capacity-based economics and diseconomics of scope
Processor allocation for horizontal and vertical parallelism and related speedup bounds