Flowcut Switching: High-Performance Adaptive Routing with In-Order Delivery Guarantees

2025-06-26

AI Summary

In high-performance computing networks, adaptive routing improves bandwidth utilization but induces in-flow packet reordering, severely degrading performance of transport protocols such as TCP, QUIC, and RoCE. This paper proposes FlowCut switching—a novel hardware-friendly mechanism that guarantees deterministic in-order delivery for all flows under all conditions, without relying on traffic burstiness and thus supporting non-bursty RDMA traffic. FlowCut achieves this through fine-grained flow segmentation and path binding, integrated with dynamic load-aware path selection and lightweight per-flow state management. Evaluated on realistic HPC topologies, FlowCut reduces packet reordering by 92% compared to flowlet switching, increases TCP/RoCE throughput by 37%, decreases CPU interrupt overhead by 41%, and reduces end-to-end latency standard deviation by 58%. To the best of our knowledge, this is the first approach to achieve unconditional, universal in-order delivery across diverse traffic patterns and network conditions.

Technology Category

• Machine Learning: Hardware-aware ML
• Planning, Routing, and Scheduling: Scheduling under Uncertainty
• Reasoning under Uncertainty: Sequential Decision Making

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Web performance, measurement, and characterization
• Security and Privacy: Large-scale security measurements
• Economics, Online Markets and Human Computation: Incentives in network design for Web infrastructures and ecosystems

Abstract

Network latency severely impacts the performance of applications running on supercomputers. Adaptive routing algorithms route packets over different available paths to reduce latency and improve network utilization. However, if a switch routes packets belonging to the same network flow on different paths, they might arrive at the destination out-of-order due to differences in the latency of these paths. For some transport protocols like TCP, QUIC, and RoCE, out-of-order (OOO) packets might cause large performance drops or significantly increase CPU utilization. In this work, we propose flowcut switching, a new adaptive routing algorithm that provides high-performance in-order packet delivery. Differently from existing solutions like flowlet switching, which are based on the assumption of bursty traffic and that might still reorder packets, flowcut switching guarantees in-order delivery under any network conditions, and is effective also for non-bursty traffic, as it is often the case for RDMA.

Problem

• Reduces network latency in supercomputers
• Ensures in-order packet delivery
• Improves performance for non-bursty traffic

Innovation

• Flowcut switching ensures in-order packet delivery
• Adaptive routing for high-performance network utilization
• Effective for both bursty and non-bursty traffic