Performance analysis of practical distributed backpressure protocols

Backpressure Routing is a cross-layer control algorithm that makes decisions on routing and resource allocation in mobile ad hoc networks using network state information like queue backlog values and current available channel rates. While it is desirable because of its provable performance guarantees, including providing optimal throughput, the real world problem of implementing backpressure routing in a distributed fashion is often ignored. To gain insight into how scheduling decisions and their impacts on performance are affected by errors, we develop two practical protocols that exchange network state information to allow nodes to make scheduling decisions in a distributed fashion. One protocol represents using the most current network state information, even if that information is inconsistent across different nodes. The second protocol models the situation in which nodes use information that is consistent, but is not current, and therefore, highly likely to contain errors compared to the current network state. We analyze the possible scheduling outcomes, modeling the probability of each, which allows us to compare the protocols´ performance. We also provide results from simulations implemented in ns-3 that support the analysis. With this analysis, we are able to show that small magnitude errors have a large impact on performance that is abated as the error grows. We are also able to show the benefits of using consistent information in scheduling, since it fully utilizes available channels.