A Scalable Architecture for Distributed Shared Memory Multiprocessors Using Optical Interconnects

In this paper, we describe the design and analysis of a scalable architecture suitable for large-scale DSMs (Dis-tributed Shared Memory) systems. The approach is based on an interconnect technology which combines optical com-ponents and a novel architecture design. In DSM sys-tems, as the network size increases, network contention results in increasing the critical remote memory access latency, which significantly penalizes the performance of DSM systems. In our proposed architecture called RAPID (Reconfigurable and scalable All-Photonic Interconnect for Distributed-shared memory), we provide high connectiv-ity by maximizing the channel availability for remote com-munication to reduce the remote memory access latency. RAPID also provides fast and efficient unicast, multicast and broadcast capabilities using a combination of aggres-sively designed wavelength, time and space-division multi-plexing techniques. We evaluated RAPID based on network characteristics, power budget criteria and simulation using synthetic traffic workloads and compared it against other scalable electrical networks. We found that RAPID, not only outperforms other networks, but also, satisfies most of the requirements of shared memory multiprocessor design such as low latency, high bandwidth, high connectivity, and easy scalability.