IEEE Network · 2024 · 43 citations · 17 references
ComputingComputer EngineeringComputer ScienceCloud InfrastructureCloud NetworkingComputer NetworksCongestion ControlLatency OptimizationRemote Memory AccessTransport LayerCongestion Control ProtocolsDatacenter Congestion ControlHigh-performance NetworkingKernel Bypass NetworkingPerformance RequirementsRdma Transport ProtocolsRemote Direct Memory AccessScalability
Remote Direct Memory Access (RDMA) has become an important building block of modern datacenter network (DCN) infrastructure given the merits of kernel bypass, zero memory copy techniques and low CPU overhead. However, the increasingly stringent performance requirements of both ultra-low latency and high throughput from the booming datacenter applications raise substantial challenges to the widespread adoption of RDMA in large-scale DCNs. This article identifies and discusses specific challenges facing RDMA deployment, including congestion control, multi-path transmission and scalability issues. We provide a comprehensive review of the efforts undertaken over the past decade to enhance RDMA performance, highlighting the key innovations of these RDMA transports. Furthermore, we explore the insights behind these RDMA transport schemes and outline the future research opportunities with the hope to inspire new interests and shed light on the following research in this field.
17
Mohammad Alizadeh, Tom Edsall, Sarang Dharmapurikar et al. · 2014 · 701 citations
Yuliang Li, Rui Miao, Hongqiang Harry Liu et al. · 2019 · 594 citations
RDMA over Commodity Ethernet at Scale
Chuanxiong Guo, Haitao Wu, Zhong Deng et al. · 2016 · 454 citations
Congestion Control for Large-Scale RDMA Deployments
Yibo Zhu, Haggai Eran, Daniel Firestone et al. · 2015 · 392 citations · Full text
Radhika Mittal, Vinh The Lam, Nandita Dukkipati et al. · ACM SIGCOMM Computer Communication Review · 2015 · 315 citations · Full text