Stochastic network models for ATM networks / William Liang.

Format:

Book

Manuscript

Microformat

Thesis/Dissertation

Author/Creator:

Liang, William.

Contributor:

Ross, Keith W., 1956- advisor.

University of Pennsylvania.

Language:

English

Subjects (All):

Penn dissertations--Systems engineering.

Systems engineering--Penn dissertations.

Local Subjects:

Penn dissertations--Systems engineering.

Systems engineering--Penn dissertations.

Physical Description:

xi, 122 pages : illustrations ; 29 cm

Production:

1997.

Summary:

For bursty sources, statistical Multiplexing in ATM networks leads to better utilization of network resources than does peak rate admission. In this dissertation we investigate several ATM admission control policies which are based on statistically multiplexing cells at the input to each link in an ATM network. The admission policies under investigation differ by whether the cells are segregated or multiplexed across service types. For the case where the services are separately multiplexed, the allocation of bandwidth for services can be either static or dynamic. In the network context, admission policies are further differentiated by the degree of multiplexing permitted among the VCs with different routes. For each of the admission policies we determine explicit expressions for the VC admission region and show how VC blocking can be obtained from a product-form expression. These admission regions involve nonlinear inequality constraints, thus generalizing the admission regions for multiservice loss networks with peak rate admission. We then apply the theory to a three-link network with three service types: voice, video conference, and LAN interconnect data. We show that Monte Carlo summation is an effective tool for computing VC blocking probabilities resulting from the different admission policies, which enables us to evaluate and compare the various admission policies. We also show that reduced load approximation can be applied to some of the admission policies to approximate VC blocking probabilities when used in conjunction with a heuristic to derive a conservative linear admission region. Using the results obtained from the numerical example as a guide, we discuss the tradeoffs between implementation complexity and VC connection performance of the various admission policies.

Notes:

Supervisor: Keith W. Ross.

Thesis (Ph.D. in Systems Engineering) -- University of Pennsylvania, 1997.

Includes bibliographical references.

Local Notes:

University Microfilms order no.: 98-00891.

OCLC:

244970913