Managing End-to-End Network Performance via Optimized Monitoring Strategies

  • Authors:
  • H. Cenk Ozmutlu;Natarajan Gautam;Russell Barton

  • Affiliations:
  • Department of Industrial Engineering, School of Engineering and Architecture, Uludag University, Gorukle, Bursa, 16059, Turkey/ hco@uludag.edu.tr;Department of Industrial and Manufacturing Engineering, Pennsylvania State University, University Park, Pennsylvania 16802/ ngautam@psu.edu;Department of Industrial and Manufacturing Engineering, Pennsylvania State University, University Park, Pennsylvania 16802/ rbarton@psu.edu

  • Venue:
  • Journal of Network and Systems Management
  • Year:
  • 2002

Quantified Score

Hi-index 0.00

Visualization

Abstract

To predict the delay between a source and a destination as well as to identify anomalies in a network, it is useful to continuously monitor the network by sending probes between all sources and destinations. Some of the problems of such probing strategies are: (1) there is a very large amount of information to analyze in real time; and (2) the probes themselves could add to the congestion. Therefore it is of prime importance to reduce the number of probes drastically and yet be able to reasonably predict delays and identify anomalies. In this paper we formulate a graph-theoretic problem called the Constrained Coverage Problem to optimally select a subset to traceroute-type probes to monitor networks where the topology is known. To solve this problem, we develop a heuristic algorithm called the Constrained Coverage Heuristic (CCH) algorithm, which works in polynomial time, as an alternative to the standard exponential-time integer programming solution available in commercial software. The application of the Constrained Coverage Problem to randomly generated topologies yielded an 88.1% reduction in the number of monitored traceroute-type probes on average. In other words, networks can be successfully monitored using only 11.9% of all possible probes. For these examples, the polynomial time CCH algorithm performed remarkably well in comparison to the standard exponential time integer programming algorithm and obtained the optimal (in 24 of 30 examples) or near optimal (second best solution in the remaining examples) solutions in a comparatively negligible amount of time.