ProFlow: RL-Driven and Performance-Aware Proactive Flow Placement in Datacenter Networks

2026-07-28

AI Summary

In multi-tenant data center networks, competing flows sharing aggregation switches often induce congestion that degrades the throughput of performance-sensitive flows. To address this challenge, this work proposes ProFlow, a novel framework that leverages distributed telemetry to detect early signs of congestion and proactively reroutes protected flows using an offline-trained reinforcement learning model, thereby enabling performance-aware, proactive flow placement. By tightly integrating reinforcement learning with real-time network telemetry, ProFlow achieves approximately 40% higher average throughput compared to reactive rerouting baselines on the FABRIC testbed, while triggering rerouting decisions an average of 34 seconds earlier.

Abstract

In datacenter fabrics composed of leaf and aggregation switches, competing flows may become co-located on shared aggregation switches, creating congestion that can significantly degrade protected flows. However, before throughput degradation becomes observable, the network often exhibits early signs characterized by rising flow activity and queue overflow signals. Existing congestion-management approaches primarily react only after congestion becomes visible, leaving these early signs largely unexploited. In this paper, we propose ProFlow, a proactive flow-placement framework for protecting performance-sensitive traffic in multi-tenant datacenter networks, thereby utilizing the early signs of potential throughput degradations. ProFlow leverages distributed telemetry signals and offline-trained reinforcement learning (RL) to identify precursor congestion conditions and proactively reroute protected flows before throughput degradation occurs. Evaluation results using FABRIC testbed show that ProFlow achieves approximately 40% higher mean throughput than a reactive rerouting baseline while initiating rerouting decisions around 34 seconds earlier on average, demonstrating the effectiveness of anticipatory congestion management.

Problem

• congestion management
• flow placement
• datacenter networks
• throughput degradation
• proactive rerouting

Innovation

• proactive flow placement
• reinforcement learning
• congestion prediction
• performance-aware routing
• datacenter networks