Multihop networks: Performance modeling under non-uniform traffic patterns

Performance modeling under non-uniform traffic is a useful tool to validate simulation accuracy and lend insights to realistic implementation of multihop networks. In this article, we present a performance model capable of tracking non-uniform traffic for an arbitrary multihop network. Our model is a an extension of Greenberg-Goodman and Brassil-Cruz models which are limited to Manhattan Street networks. By considering packets with non-null states only, we also provide an improved computational efficiency. Application to estimate the performance of Toroidal Mesh and Diagonal Mesh networks under non-uniform traffic is provided. Comparison of performance parameters derived from the model and from simulation show good agreement. Limitations of our memory-less and independence assumptions based model are addressed.