Energy-efficiency of load-balancing routing for wireless convergecast networks: Centralized versus distributed implementation

In a wireless convergecast network, load-balancing routing contributes to improving energy-efficiency by eliminating the difference of relaying load among one-hop nodes. Load-balancing algorithms can be operated by centralized and distributed implementation. Centralized implementation needs to collect global neighborhood information, which leads to a large amount of signaling overhead. On the contrary, distributed implementation can overcome such issue. However, it only achieves sub-optimal load-balancing. Thus, it is obscure that which type of implementation is more energy-efficiency. In this paper, a novel distributed implementation for load-balancing routing is proposed. We make a series of analyses of comparing the energy-efficiency between centralized and distributed implementation. The results show that, the overhead for constructing load-balancing routing via distributed implementation is always less than that via centralized implementation. However, the heaviest one-hop node in distributed implementation suffers from more loads so that according to its energy consumption, distributed implementation can maintain energy-efficiency only in the condition that routing update cycle is shorter than a threshold.