Probabilistic Breadth-First Search – A Method for Evaluation of Network-Wide Broadcast Protocols
Tuesday, January 21, 2014, 10:46Authors: Björn Lichtblau Andreas Dittrich 6th IEEE/ACM/IFIP International Conference on New Technologies, Mobility and Security, NTMS 2014, Dubai, UAE, March 30 – April 2, 2014 Download: accepted version, final published version |
In wireless mesh networks (WMNs), network-wide broadcasts (NWBs) are a fundamental operation, required by routing and other mechanisms that distribute information to all nodes in the network. However, due to the characteristics of wireless communication, NWBs are generally problematic. Optimizing them is thus a prime target when improving the overall performance and dependability of WMNs. Most of the existing optimizations neglect the real nature of WMNs and are based on simple graph models, which provide optimistic assumptions of NWB dissemination. On the other hand, models that fully consider the complex propagation characteristics of NWBs quickly become unsolvable due to their complexity. In this paper, we present the Monte Carlo method probabilistic breadth-first search (PBFS) to approximate the reachability of NWB protocols. PBFS simulates individual NWBs on graphs with probabilistic edge weights, which reflect link qualities of individual wireless links in the WMN, and estimates reachability over a configurable number of simulated runs. This approach is not only more efficient than existing ones, but further provides additional information such as the distribution of path lengths. Furthermore, it is easily extensible to NWB schemes other than flooding. The applicability of PBFS is validated both theoretically and empirically, in the latter by comparing reachability as calculated by PBFS and measured in a real-world WMN. Validation shows that PBFS quickly converges to the theoretically correct value and approximates the behaviour of real-life testbeds very well. The feasibility of PBFS to support research on NWB optimizations or higher level protocols that employ NWBs is demonstrated in two use cases.
Categories: Publication, Research and Education
Tags: Distributed computing, Monte Carlo methods, Network, Network-wide broadcasts, Probabilistic network graphs, Wireless mesh networks