mPlane: an architecture for scalable fault localization

  • Authors:

  • Ramana Rao Kompella;Alex C. Snoeren;George Varghese

  • Affiliations:

  • Purdue University, West Lafayette, IN;University of California, San Diego, La Jolla, CA;University of California, San Diego, La Jolla, CA, USA

  • Venue:
  • Proceedings of the 2009 workshop on Re-architecting the internet
  • Year:
  • 2009

Quantified Score

Hi-index 0.00

Visualization

Abstract

Customers are increasingly demanding service-level guarantees from ISPs. Unfortunately, ISPs today have limited ability to measure and monitor their own networks. ISPs use active probes for monitoring network health and use tomographic approaches to localize any end-to-end problems observed, which are typically postulated as underconstrained problems, and hence, often limited in accuracy. Active probes are also fundamentally unscalable; operators cannot afford to inject them at high frequencies. We present an architecture, mPlane, that addresses these problems. The key idea in mPlane is to break paths into segments consisting of router forwarding paths and links, and conduct measurements on a per-segment basis. Node measurements are obtained through scalable high-fidelity hardware primitives, while the link measurements are conducted using segment-level active probes at low frequencies. Unfortunately, upgrading all routers with these primitives faces significant deployment hurdles; we propose an incremental deployment strategy that picks the most important routers to upgrade. Our simulations with RocketFuel topologies indicate a partial deployment on 15% of an ISP's routers can yield two orders of magnitude decrease in measurement overhead, and reduce the average localization granularity from 4 hops to about 1.5.