Balancing overhead-minimization objectives in network probing-path selection

Abstract

Probing is a technique used for network monitoring and measurement applications, including detection of silent failures and performance analysis to verify compliance with service-level agreements. Many existing approaches to probing-path selection seek to minimize the number of probes (an NP-hard problem) or to minimize probing delay, but they fail to properly assess the network overhead imposed by the probing strategy. In this paper, we present an approximation algorithm for probing-path selection that simultaneously considers two types of overhead-minimization objectives. Our approach can be customized to achieve a desired balance between the number of probes used overall and the effect on individual network components such as network devices and links. Through analysis and simulation on various topologies and under a variety of settings, we assess the effect of customization on a number of network-overhead measures. Compared to previous work, our approach efficiently computes a set of end-to-end probing paths fully covering a network without unduly overloading it.