Resource Allocation Issues in Communication Networks

 

Tara Javidi

PhD Candidate

University of Michigan

Department of Electrical Engineering and Computer Science

 

 

Rapid growth in size of communication networks and the complexity of applications and offered services provide numerous challenges in the design and study of communication systems. Resource allocation is a major component of network utilization and quality of service provisioning in such systems. Because of scarcity of resources and the demand for better quality of service, optimal resource allocation problems play an ever-increasing role in the design and analysis of communication networks. In this talk the merits and elements of an analytic approach to such problems are discussed, first. Then, a few examples of resource allocation issues in networking, from connection admission control in cellular systems to scheduling algorithms in high speed routers to routing in ad hoc wireless networks, are introduced briefly. Finally the following connection admission problem in wireless networks is presented.

 

The mobile wireless environment, in particular, provides serious challenges such as limited bandwidth, low capacity channels, and interference among users. As a result, an important network layer problem in the design of cellular systems is how to allocate the limited resources (e.g.\ bandwidth) efficiently, while providing quality of service (QoS) guarantees to the applications in terms of bit-rate and loss. This problem becomes more acute for the next-generation integrated-services networks which aim to support heterogenous traffic which aim to support heterogenous traffic. In cellular networks desirable resource allocation is mainly achieved by the base station through connection admission decisionsin the presence of new connection requests. Granting an admission is equivalent to a contract where the newly connected service is guaranteed a set of desirable QoS measures for the full length of the connection, while generating revenue at a prespecified rate. As a result, an efficient allocation of resources is achieved by constructing a connection admission strategy that maximizes the average expected revenue and satisfies the QoS requirements of each connected service. In this talk it is shown that, mathematically, this problem can be represented as a constrained stochastic dynamic optimization problem where the QoS requirements form a dynamic set of constraints. In general such an optimization problem is computationally and analytically intractable. As an alternative approach, we propose a decomposition (possibly suboptimal) of the problem into two tractable subproblems. Furthermore, it will be shown that such a decomposition creates a conceptual framework for understanding the interaction among various layers of the system such as physical layer concerns, QoS requirements, and network layer resource allocation. The nature of these sub-problems in a cellular system with an outage-based QoS requirement is illustrated and the results are discussed in detail.

Friday, March 29, 2002

3:30 - 5:00 p.m.

1500 EECS