Throughput in locally balanced computer system models
Conference
·
OSTI ID:7210330
The optimization of throughput in locally balanced queueing network models is investigated. A general result, useful in the design of computer system models, shows that throughput is a non-decreasing function of the number of customers contained in any subnetwork. The throughput of the two-queue network, which duplicates (through Norton's theorem) that of more complex networks, is characterized in terms of the processing rates at each queue. Then processor allocation algorithms that maximize throughput are shown for the case where processing power can be switched between queues. The maximization of throughput is shown first in the case that processing power allocations to a queue depend on the queue state only, and then, in an extension of known locally balanced queue, the case in which processing power is allocated on the basis of an entire subnetwork state. The latter case provides a simple and optimum rule for processor allocations that maximize throughput in networks containing multiprocessor systems. 2 figures.
- Research Organization:
- Brookhaven National Lab., Upton, NY (USA)
- DOE Contract Number:
- EY-76-C-02-0016
- OSTI ID:
- 7210330
- Report Number(s):
- BNL-22531; CONF-770812-1
- Country of Publication:
- United States
- Language:
- English
Similar Records
Throughput in locally balanced computer system models. [Optimization]
Throughput in locally balanced computer system models
Conditions for product form network solution through the analysis of finite queues with overflows
Technical Report
·
Fri Dec 31 23:00:00 EST 1976
·
OSTI ID:7299883
Throughput in locally balanced computer system models
Technical Report
·
Wed Jun 01 00:00:00 EDT 1977
·
OSTI ID:5663111
Conditions for product form network solution through the analysis of finite queues with overflows
Conference
·
Sat Dec 31 23:00:00 EST 1977
·
OSTI ID:6750085