Can we schedule traffic more efficiently in optical packet switches?

We consider traffic scheduling in non-blocking electronic-buffered optical packet switches (OPS) with bounded packet delay. Due to the reconfiguration overhead of the switch fabric, the two commonly used optimization objectives, minimizing packet delay and minimizing switch speedup, conflict with each other. Intelligent scheduling algorithms have been designed to provide tradeoff between these two objectives. In this paper, we propose a more efficient approach to schedule OPS traffic, resulting in significantly reduced speedup and/or packet delay. However, our approach is based on a very interesting conjecture, which has not been strictly proved so far. We would like to put forward this conjecture as an open question, and call for a proof or disproof.