PBProbe: A capacity estimation tool for high speed networks

  • Authors:

  • Ling-Jyh Chen;Tony Sun;Bo-Chun Wang;M. Y. Sanadidi;Mario Gerla

  • Affiliations:

  • Institute of Information Science, Academia Sinica, 128, Sec. 2, Academia Road, Taipei 11529, Taiwan;PacketMotion Inc., 110 Baytech Drive, 2nd Floor San Jose, CA 95134, USA;Institute of Information Science, Academia Sinica, 128, Sec. 2, Academia Road, Taipei 11529, Taiwan;Department of Computer Science, University of California at Los Angeles 4732 Boelter Hall, Los Angeles, CA 90095, USA;Department of Computer Science, University of California at Los Angeles 4732 Boelter Hall, Los Angeles, CA 90095, USA

  • Venue:
  • Computer Communications
  • Year:
  • 2008

Quantified Score

Hi-index 0.24

Visualization

Abstract

Knowledge about the bottleneck capacity of an Internet path is critical for efficient network design, management, and usage. In this paper, we propose a new technique, called PBProbe, for estimating high speed links rapidly and accurately. Although it is based on CapProbe, instead of relying solely on packet pairs, PBProbe employs the concept of ''Packet Bulk'' and adapts the bulk length to compensate for the well known problem with packet pair-based approaches, namely the lack of accurate timer granularity. As a result, PBProbe not only preserves the simplicity and speed of CapProbe, but also correctly estimates link capacities over a much larger range. Using analysis, we evaluate PBProbe with various bulk lengths, network configurations, and traffic models. We then perform a set of experiments to evaluate the accuracy of PBProbe on the Internet over wired and wireless links. Finally, we perform emulation and Internet experiments to verify the accuracy and speed of PBProbe on high speed links (e.g., the Gigabit Ethernet connection). The results show that PBProbe is consistently fast and accurate in the majority of test cases.