DESIGN AND EVALUATION OF A METHOD FOR ADMISSION CONTROL FOR BROAD-BAND PACKET NETWORKS

Citation
Wp. Yeh et al., DESIGN AND EVALUATION OF A METHOD FOR ADMISSION CONTROL FOR BROAD-BAND PACKET NETWORKS, Computer networks and ISDN systems, 27(9), 1995, pp. 1349-1366
Citations number
26
Categorie Soggetti
Computer Sciences","System Science",Telecommunications,"Engineering, Eletrical & Electronic","Computer Science Information Systems
ISSN journal
01697552
Volume
27
Issue
9
Year of publication
1995
Pages
1349 - 1366
Database
ISI
SICI code
0169-7552(1995)27:9<1349:DAEOAM>2.0.ZU;2-9
Abstract
In this paper two admission control algorithms are defined and evaluat ed. The algorithms are designed for use with an architecture which seg regates traffic flows into multiple queues, one for each individual co nnection or one for each class of connections. The approach makes it p ossible to provide multiple classes of service. Resource allocation ta bles can be prepared efficiently, using simulation. The resulting tabl es are independent of possible traffic mixtures and require only two s calar numbers to specify required resources. The admission process can be carried out by passing the two numbers between nodes, and it is no t necessary to provide detailed information about traffic mixtures. Th e paper presents a comprehensive performance study, carried out with s imulation, to evaluate the performance of the proposed admission contr ol algorithms in relation to other algorithms described in the literat ure. Under the conditions used in the performance study, simulation is used to establish an admissible region of all possible combinations o f the numbers of sources in two classes of traffic which can be suppor ted by a specific channel independent of the admission algorithm. Thus , the algorithms are compared to the best that is possible as well as competing algorithms. For the case of multiple sources with different quality of service requirements, one of the proposed algorithms outper formed three competing algorithms and gave results close to the best t hat can be achieved with any admission algorithm.