提出最优负载均衡方案，显著提升AI训练网络性能。
Load Balancing for AI Training Workloads
- 采用基于包喷洒的负载均衡，优于传统流级调度
- 主机端方案在链路故障时表现更优，响应更快
- 提出目标旋转机制，实现理想队列缩放，适合高并发场景
AI训练对带宽的极端需求使负载均衡成为AI网络架构中的关键组件，近年来产业界和学术界涌现出多种设计方案。然而，目前尚无共识明确哪种方法占优，或其适用条件。我们缺乏对现有方法与最优性能之间差距的理解。为此，我们系统评估了主流负载均衡设计，将其与具体拥塞控制和丢包恢复机制解耦。结果表明，基于包喷洒的负载均衡优于在流、流段或子流粒度上的传统方法。在主机端与交换机端的包喷洒方案对比中，两者在无故障情况下表现相近，但主机端在链路故障时因能快速感知端到端路径状态而更具优势。我们还发现，现有任何主流方案均无法在最大利用率下实现最优的O(1)队列缩放。我们证明了基于目标旋转（DR）的调度策略可达到该最优，并提出Ofan——一种交换机端实现的DR方案，实验显示其性能显著优于其他包喷洒方法。
原文摘要 · Abstract (English)
The extreme bandwidth demands of AI training has made load-balancing a critical component in AI fabrics, and a variety of load-balancing designs have emerged in recent work from both industry and research. However, there is currently little consensus on which design approach dominates or the conditions under which an approach dominates. We also lack an understanding of how far these approaches are from optimal. We provide a technical foundation for answering these questions by systematically evaluating leading load-balancing designs, while decoupling them from specific congestion control and loss recovery stacks. We find that load-balancing based on packet spraying dominates traditional approaches that load balance traffic at flow, flowlet, or subflow granularities. When comparing host- vs switch-based approaches to packet spraying, we find that they perform similarly in failure-free scenarios but that a host-based approach dominates under link failure because of its rapid visibility into end-to-end path conditions. We also identify that no leading approach achieves optimal O(1) queue scaling at maximum utilization. We demonstrate why a destination-based rotation (DR) discipline can reach this optimum and introduce Ofan, a switch-based implementation of DR that we show offers valuable performance gains over other packet spraying approaches.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。