Some studies on the impact of dynamic traffic in a QoS-based dynamic routing environment

We consider dynamic QoS routing in an integrated-services environment where some of the offered traffic streams are dynamic. While there has been work on routing performance in the presence of purely stationary traffic, the presence of dynamic traffic has received very little attention. We consider three routing schemes: dynamic random routing (DRR), maximum available capacity routing with periodic update and crankback (MACRPC) and maximum available capacity routing with instant computations (MACRIC). Based on simulation, we have found that while DRR performs and adapts better than MACRPC in the case of no trunk reservation, it is other way around when there is moderately high trunk reservation (even for less frequent routing updates). Further, we found that dynamics of a traffic class can result in dynamic flow blocking behavior even on stationary traffic, especially at a low or no trunk reservation level