Optically connected memory for disaggregated data centers

• Published in: Journal of parallel and distributed computing 2022-05, Vol.163 (C), p.300-312
• Main Authors: Gonzalez, Jorge, G. Palma, Mauricio, Hattink, Maarten, Rubio-Noriega, Ruth, Orosa, Lois, Mutlu, Onur, Bergman, Keren, Azevedo, Rodolfo
• Format: Article
• Language: English
• Subjects: Data-centers; Disaggregated memory; DRAM; MATHEMATICS AND COMPUTING; Memory systems; Photonics
• ISSN: 0743-7315; 1096-0848
• DOI: 10.1016/j.jpdc.2022.01.013

•Optical memory disaggregation for data centers achieves low energy-per-bit consumption.•Main memory disaggregation using state-of-the-art optical devices.•First evaluation of both energy-per-bit and system-level performance for main memory disaggregation with optical devices. Recent advances in integrated photonics enable the implementation of reconfigurable, high-bandwidth, and low energy-per-bit interconnects in next-generation data centers. We propose and evaluate an Optically Connected Memory (OCM) architecture that disaggregates the main memory from the computation nodes in data centers. OCM is based on micro-ring resonators (MRRs), and it does not require any modification to the DRAM memory modules. We calculate energy consumption from real photonic devices and integrate them into a system simulator to evaluate performance. Our results show that (1) OCM is capable of interconnecting four DDR4 memory channels to a computing node using two fibers with 1.02 pJ energy-per-bit consumption and (2) OCM performs up to 5.5× faster than a disaggregated memory with 40G PCIe NIC connectors to computing nodes.