Using adaptive routing to achieve Quality of Service

Self-monitoring allows a network to observe its own behavior via probing and measurement mechanisms. This can then be exploited by the system to take autonomous decisions for the purpose of system management, performance management and user Quality of Service (QoS). In this paper, we experimentally explore how QoS goals that are externally set by network users, can then be explicitly exploited by a self-aware network to control its own behavior to attain these goals. The experiments we report are conducted in two distinct Cognitive Packet Network (CPN) test-beds which use probing to select the routes which best satisfy the QoS goal. Our experiments validate this concept with QoS goals which include end-to-end delay, packet loss, and a mixture of these two metrics. We observe that using only delay in the QoS goal is a good way to reduce delay and loss if losses result only from congestion. However, using loss in the QoS goal is seen to be useful if the paths that are adaptively selected avoid nodes where packet losses occur for reasons other than congestion. In general we observe that CPN networks effectively adapt routing behavior to the QoS goal that is specified.