Bandwidth-efficiency-oriented topology optimization for integrated switching systems based on circulant graphs

This work studies a topology optimization problem by leveraging a regular network topology-circulant graph. The result of the topology optimization is applied to the design of the integration network for modern integrated switching systems (ISSs), which integrate individual switching devices together to boost switching capacity and to reduce network management complexity. The target of the optimization is to maximize bandwidth efficiency on the integration links and hence to maximize the ISS capacity to accommodate traffic. This eventually translates to minimizing the average inter-node hop distance within the ISS.First, the problem is formulated into an integer linear program (ILP) and it is solved to its optimality for moderate-size networks. Then, by comparing with the Moore bound and unconstrained topology optimization, we show the circulant graph can be a qualified candidate with good bandwidth efficiency. Finally, due to the scaling limitation of the ILP approach, a dynamic programming (polynomial-time) algorithm is proposed and results show that it performs very closely to the optimal solutions induced from the ILP solver.