Hybrid Scheduling in High-Speed Packet Switches

The crossbar fabric switching architecture is popular due to its low cost and scalability. However, crossbar fabric switches require centralized and highly complex scheduling. Fully buffered crossbars overcome the scheduling complexity, by means of distributed scheduling. However, the scheduling simplicity comes at the expense of a complex fabric, where internal buffering is required in each crosspoint. This article proposes a novel architecture, namely the hybrid buffered crossbar (HBC) switching architecture, where only few internal buffers are maintained per output. Our goal is to design a HBC switch having the performance of buffered crossbars and the cost of unbuffered crossbars. We propose a round robin scheduling algorithm for the HBC switching architecture. Our algorithm is a combination of unbuffered as well buffered crossbar scheduling. Simulation results show interesting trade offs in choosing the optimal number of internal buffers when designing a hybrid crossbar switch.