Jupiter evolving: transforming google's datacenter network via optical circuit switches and software-defined networking
Leon Poutievski, Omid Mashayekhi, Joon Ong, Arjun Singh, Muhammad Mukarram Bin Tariq, Rui Wang, Jianan Zhang, Virginia Beauregard, Patrick Conner, Steve D. Gribble, Rishi Kapoor, Stephen Kratzer
摘要
We present a decade of evolution and production experience with Jupiter datacenter network fabrics. In this period Jupiter has delivered 5x higher speed and capacity, 30% reduction in capex, 41% reduction in power, incremental deployment and technology refresh all while serving live production traffic. A key enabler for these improvements is evolving Jupiter from a Clos to a direct-connect topology among the machine aggregation blocks. Critical architectural changes for this include: A datacenter interconnection layer employing Micro-Electro-Mechanical Systems (MEMS) based Optical Circuit Switches (OCSes) to enable dynamic topology reconfiguration, centralized Software-Defined Networking (SDN) control for traffic engineering, and automated network operations for incremental capacity delivery and topology engineering. We show that the combination of traffic and topology engineering on direct-connect fabrics achieves similar throughput as Clos fabrics for our production traffic patterns. We also optimize for path lengths: 60% of the traffic takes direct path from source to destination aggregation blocks, while the remaining transits one additional block, achieving an average block-level path length of 1.4 in our fleet today. OCS also achieves 3x faster fabric reconfiguration compared to pre-evolution Clos fabrics that used a patch panel based interconnect.
问问这篇 Paper
智能体会读完全文。
Lune 把这篇 Paper 索引到了每一个公式,引用它的顶会 Paper 也一样。你提问,回答直接引用原文。
引用它的顶会 Paper38
- TopoOpt: Co-optimizing Network Topology and Parallelization Strategy for Distributed Training JobsWeiyang Wang, Moein Khazraee, Zhizhen Zhong, Manya Ghobadi 等NSDI 2023 · 被引用 215 次
- Alibaba HPN: A Data Center Network for Large Language Model TrainingKun Qian, Yongqing Xi, Jiamin Cao, Jiaqi Gao 等SIGCOMM 2024 · 被引用 173 次
- Empowering Azure Storage with RDMAWei Bai, Shanim Sainul Abdeen, Ankit Agrawal, Krishan Kumar Attre 等NSDI 2023 · 被引用 117 次
- Resiliency at Scale: Managing Google's TPUv4 Machine Learning SupercomputerYazhou Zu, Alireza Ghaffarkhah, Hoang-Vu Dang, Brian Towles 等NSDI 2024 · 被引用 46 次
- FIGRET: Fine-Grained Robustness-Enhanced Traffic EngineeringXimeng Liu, Shizhen Zhao, Yong Cui, Xinbing WangSIGCOMM 2024 · 被引用 36 次
它引用的顶会 Paper6
- Swift: Delay is Simple and Effective for Congestion Control in the DatacenterGautam Kumar, Nandita Dukkipati, Keon Jang, Hassan M. G. Wassel 等SIGCOMM 2020 · 被引用 333 次
- Sirius: A Flat Datacenter Network with Nanosecond Optical SwitchingHitesh Ballani, Paolo Costa, Raphael Behrendt, Daniel Cletheroe 等SIGCOMM 2020 · 被引用 204 次
- Expanding across time to deliver bandwidth efficiency and low latencyWilliam M. Mellette, Rajdeep Das, Yibo Guo, Rob McGuinness 等NSDI 2020 · 被引用 194 次
- Orion: Google's Software-Defined Networking Control PlaneAndrew D. Ferguson, Steve D. Gribble, Chi-Yao Hong, Charles Killian 等NSDI 2021 · 被引用 95 次
- Experiences with Modeling Network Topologies at Multiple Levels of AbstractionJeffrey C. Mogul, Drago Goricanec, Martin Pool, Anees Shaikh 等NSDI 2020 · 被引用 43 次
相关 Paper
- Rearchitecting Datacenter Networks: A New Paradigm with Optical Core and Optical EdgeSushovan Das, Arlei Silva, T. S. Eugene NgINFOCOM 2024 · 被引用 5 次
- Cost-Effective Live Expansion of Three-Stage Switching Networks without Blocking or Connection RearrangementTakeru Inoue, Toru Mano, Takeaki UnoINFOCOM 2023 · 被引用 6 次
- Lightwave Fabrics: At-Scale Optical Circuit Switching for Datacenter and Machine Learning SystemsHong Liu, Ryohei Urata, Kevin Yasumura, Xiang Zhou 等SIGCOMM 2023 · 被引用 63 次
- ATRO: A Fast Algorithm for Topology Engineering of Reconfigurable Datacenter NetworksYingming Mao, Qiaozhu Zhai, Ximeng Liu, Xinchi Han 等INFOCOM 2026 · 被引用 3 次
- NegotiaToR: Towards A Simple Yet Effective On-demand Reconfigurable Datacenter NetworkCong Liang, Xiangli Song, Jing Cheng, Mowei Wang 等SIGCOMM 2024 · 被引用 27 次
