Quality of Service Testing Methodology

Typically, a network service provider offers best-effort service to customers and Quality of Service (QoS) is provided only by network over provisioning. When traffic bursts cause congestion, this default approach does not always provide an acceptable solution. As the trend in networking is convergence of voice, video and data, it is important to differentiate traffic into high and low priority classes through QoS mechanisms. To deliver differentiated QoS, the routers utilize classification and sometimes remarking of the traffic. Based on service agreements with the customer, a certain class of traffic is then given more resources by the router. Evaluation of the results of these methods and approaches on the network performance is very important to device designers, network evaluators, network administrators, network operators and others involved. To study the effects of QoS solutions/techniques on the traffic, methods for testing need to be developed. The objective of this thesis is to propose methods of measuring QoS metrics like throughput, latency, and jitter for different types of traffic through a device. Classification and remarking were observed to put extra load on a device. For some devices, the performance deteriorates when classification and remarking are enabled. Throughput was seen to drop for classification and remarking in some cases. In multiple instances, latency and jitter increased for classification and remarking. There are multiple ways of measuring jitter which were also discussed in the methodology development. It was also observed that some definitions which might seem better from a research perspective are not always available to measure or widely used in the industry. Based on the study it was concluded that service/network providers will have to take care while providing QoS to the customers in order avoid the scenarios discussed here.

Authors: 
Ankur Chadda
Category: 
Theses