Silicon Photonic Flex-LIONS for Reconfigurable Multi-GPU Systems

• Published in: Journal of lightwave technology Vol. 39; no. 4; pp. 1212 - 1220
• Main Authors: Fariborz, Marjan, Xiao, Xian, Fotouhi, Pouya, Proietti, Roberto, Yoo, S. J. Ben
• Format: Journal Article
• Language: English
• Published: New York IEEE 15.02.2021; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Bandwidth; Connectivity; Convolution; Engineering; Graphics processing units; Microprocessors; Multi-GPU systems; Optical fiber cables; Optical fiber filters; Optical fiber networks; Optical interconnections; optical reconfiguration; Optical switches; optical switching; Optics; Parallel processing; Reconfiguration; silicon photonics; Steering; System effectiveness; Telecommunications; Traffic congestion
• ISSN: 0733-8724; 1558-2213
• DOI: 10.1109/JLT.2021.3052713

The rapid increases in data-intensive applications demand for more powerful parallel computing systems capable of parallel processing a large amount of data more efficiently and effectively. While GPU-based systems are commonly used in such parallel processing, the exponentially rising data volume can easily saturate the capacity of the largest possible GPU processor. One possible solution is to exploit multi-GPU systems. In a multi-GPU system, the main bottleneck is the interconnect, which is currently based on PCIe or NVLink technologies. In this study, we propose to optically interconnect multiple GPUs using Flex-LIONS, an optical all-to-all reconfigurable interconnect. By exploiting the multiple free spectral ranges (FSRs) of Flex-LIONS, it is possible to adapt (or steer) the inter-GPU connectivity to the traffic demands by reconfiguring the optical connectivity of one FSR while maintaining fixed all-to-all connectivity of another FSR. Simulation results show the benefits of the proposed reconfigurable bandwidth-steering interconnect solution under various traffic patterns of different applications. Execution time reductions by up to 5× have been demonstrated in this study including two applications of convolution and maxpooling .