Performance enhancement of large scale networks with heterogeneous traffic.

Abstract

Finally, these findings are applied towards improving the performance of the Differentiated Services architecture by developing a new Refined Assured Forwarding framework where heterogeneous traffic flows share the same aggregate class. The new framework requires minimal modification to the existing Diffserv routers. The efficiency of the new architecture in enhancing the performance of Diffserv is demonstrated by simulation results under different traffic scenarios.

This dissertation builds on the notion that segregating traffic with disparate characteristics into separate channels generally results in a better performance. Through a quantitative analysis, it precisely defines the number of classes and the allocation of traffic into these classes that will lead to optimal performance from a latency standpoint. Additionally, it weakens the most generally used assumption of exponential or geometric distribution of traffic service time in the integration versus segregation studies to date by including self-similarity in network traffic.

The dissertation also develops a pricing model based on resource usage in a system with segregated channels. Based on analytical results, this dissertation proposes a scheme whereby a service provider can develop compensatory and fair prices for customers with varying QoS requirements under a wide variety of ambient traffic scenarios.

This dissertation provides novel techniques for improving the Quality of Service by enhancing the performance of queue management in large scale packet switched networks with a high volume of traffic. Networks combine traffic from multiple sources which have disparate characteristics. Multiplexing such heterogeneous traffic usually results in adverse effects on the overall performance of the network.