Delay performance of backlog based random access

Abstract

Backlog-based CSMA strategies provide a popular mechanism for distributed medium access control in wireless networks. When suitably designed, such strategies offer the striking capability to match the optimal throughput performance of centralized scheduling algorithms in a wide range of scenarios. Unfortunately, however, the activation rules used in these schemes tend to yield excessive backlogs and delays. More aggressive activation rates can potentially improve the delay performance, but may not allow provable maximum-stability guarantees. In order to gain a fundamental understanding how the shape of the activation function affects the queueing behavior, we focus on a single- node scenario, thus separating the impact of the network topology. We demonstrate that three qualitatively different regimes can arise, depending on how rapidly the activation function increases with the backlog. Simulation experiments are conducted to validate the analytical findings.