Analysis of a transmission scheduling algorithm for supporting bandwidth guarantees in bufferless networks

Abstract

In a network of bufferless packet multiplexers, the userperceived capacity of an ingress-egress tunnel (connection) may degrade quickly with increasing path length. This is due to the compounding of transmission blocking probabilities along the path of the connection, even when the links are not overloaded. In such an environment, providing users (e.g., client ISPs) with tunnels of statistically guaranteed bandwidth may limit the network's connection-carrying capacity.In this paper, we introduce and analyze a transmissionscheduling algorithm that employs randomization and traffic regulation at the ingress, and batch scheduling at the links. The algorithm ensures that a fraction of transmissions from each connection is consistently subject to small blocking probability at every link, so that these transmissions are likely to survive long paths. For this algorithm, we obtain tight bounds on the expectation and tail probability of the blocking rate of any ingress-egress connection. We compare the bounds to those obtained using the FCFS link-scheduling rule. We find that the proposed scheduling algorithm significantly improves the network's connection-carrying capacity.In deriving the desired bounds, we develop an analytic framework for stochastically comparing network-wide routing and bandwidth allocation scenarios with respect to blocking in a packet multiplexer. The framework enables us to formally characterize the routing and bandwidth allocation scenarios that maximize the expected blocking rate along the path of a tagged connection.