End-to-end packet delay and loss behavior in the internet
SIGCOMM '93 Conference proceedings on Communications architectures, protocols and applications
End-to-end Internet packet dynamics
SIGCOMM '97 Proceedings of the ACM SIGCOMM '97 conference on Applications, technologies, architectures, and protocols for computer communication
Algorithms for Parameter Selection in the Weeks Method for Inverting the Laplace Transform
SIAM Journal on Scientific Computing
Hidden Markov modeling for network communication channels
Proceedings of the 2001 ACM SIGMETRICS international conference on Measurement and modeling of computer systems
On the constancy of internet path properties
IMW '01 Proceedings of the 1st ACM SIGCOMM Workshop on Internet Measurement
Fundamental bounds on the accuracy of network performance measurements
SIGMETRICS '05 Proceedings of the 2005 ACM SIGMETRICS international conference on Measurement and modeling of computer systems
Improving accuracy in end-to-end packet loss measurement
Proceedings of the 2005 conference on Applications, technologies, architectures, and protocols for computer communications
The role of PASTA in network measurement
Proceedings of the 2006 conference on Applications, technologies, architectures, and protocols for computer communications
On optimal probing for delay and loss measurement
Proceedings of the 7th ACM SIGCOMM conference on Internet measurement
An active measurement system for shared environments
Proceedings of the 7th ACM SIGCOMM conference on Internet measurement
Queen: Estimating Packet Loss Rate between Arbitrary Internet Hosts
PAM '09 Proceedings of the 10th International Conference on Passive and Active Network Measurement
| Hi-index | 0.00 |
Loss measurements are widely used in today's networks. There are existing standards and commercial products to perform these measurements. The missing element is a rigorous statistical methodology for their analysis. Indeed, most existing tools ignore the correlation between packet losses and severely underestimate the errors in the measured loss ratios. In this paper, we present a rigorous technique for analyzing performance measurements, in particular, for estimating confidence intervals of packet loss measurements. The task is challenging because Internet packet loss ratios are typically small and the packet loss process is bursty. Our approach, SAIL, is motivated by some simple observations about the mechanism of packet losses. Packet losses occur when the buffer in a switch or router fills, when there are major routing instabilities, or when the hosts are overloaded, and so we expect packet loss to proceed in episodes of loss, interspersed with periods of successful packet transmission. This can be modeled as a simple ON/OFF process, and in fact, empirical measurements suggest that an alternating renewal process is a reasonable approximation to the real underlying loss process. We use this structure to build a hidden semi-Markov model (HSMM) of the underlying loss process and, from this, to estimate both loss ratios and confidence intervals on these loss ratios. We use both simulations and a set of more than 18 000 hours of real Internet measurements (between dedicated measurement hosts, PlanetLab hosts, Web and DNS servers) to cross-validate our estimates and show that they are better than any current alternative.