International Federation For Information Processing
STRAGFLOW Tool for Straggler Detection in Data Plane
Pages
9
Time to read
41 mins
Publication
Language
English
Pages
9
Time to read
41 mins
Publication
Language
English
This document is a technical report that introduces STRAGFLOW, a tool designed for detecting stragglers in flow scheduling mechanisms within programmable data planes. It outlines the challenges posed by stragglers, which are packets that experience significantly higher queueing delays than others, leading to prolonged flow completion times (FCT). The report details how STRAGFLOW operates by monitoring queueing delays at line rate, identifying stragglers in real-time, and reporting them to the control plane. The evaluation of STRAGFLOW is conducted using real-world network traces, demonstrating its effectiveness in detecting stragglers across various scheduling schemes and link conditions. The results indicate that STRAGFLOW can provide valuable insights into straggler distribution, assisting network operators in diagnosing and mitigating flow scheduling issues to enhance overall network performance. The report also discusses the implications of straggler detection for maintaining service level agreements (SLAs) in datacenters and enterprise environments.