Impact of Network-Awareness on Profile Migration

A content-based data dissemination network routes messages to users based on their profiles. Our focus is on the design of a overlay network based filtering architecture capable of disseminating high bandwidth and high volume data streams to a large user population. This paper shows that networkaware profile assignment significantly improves the overall system throughput by jointly considering node-bound and link-bound constraints of the overlay network. Our main contribution is the design of a fully-distributed algorithm for dynamic profile movement in order to adapt to the changing network conditions. Each node in the system executes the proposed algorithm independently so that the system as a whole adapts and converges to a high-throughput operating point by efficiently utilizing the network resources. We identify the critical requirements to meet the automatic reconfiguration goal and use the insights in designing the algorithm. Based on extensive simulation study, we show that we are able to improve the system throughput by more than 80% in most of the cases and by as much as 200% in some scenarios.