A Statistical Approach to Improving Activepacket Loss Measurement
K. Siva Sankar, A. Muthukumaravel, S. Kannan · 2014
Measurement and estimation of packet loss characteristics are challenging due to the relatively rare occurrence and typically short duration of packet loss episodes. While active probe tools are commonly used to measure packet loss on end-to-end paths, there has been little analysis of the accuracy of these tools or their impact on the network. The objective of our study is to understand how to measure packet loss episodes accurately with end-to-end probes. We begin by testing the capability of standard Poisson-modulated end-to- end measurements of loss in a controlled laboratory environment using IP routers and commodity end hosts. Our tests show that loss characteristics reported from such Poisson-modulated probe tools can be quite inaccurate over a range of traffic conditions. Motivated by these observations, we introduce a sends fixed-size probes at specified intervals from new algorithm for packet loss measurement that is one measurement host to a collaborating target host. designed to overcome the deficiencies in standard The target system collects the probe packets and Poisson-based tools. Specifically, our method entails reports the loss characteristics after a specified probe experiments that follow a geometric distribution period of time. We also compare BADABING with a to 1) enable an explicit trade-off between accuracy and standard tool for loss measurement those emits impact on the network and 2) enable more accurate probe packets at Poisson intervals. The results show measurements than standard Poisson probing at the that our tool reports loss episode estimates much same rate (1-9). We evaluate the capabilities of our more accurately for the same number of probes. We also methodology experimentally by developing and show that BADABING estimates converge to the implementing a prototype tool, called BADABING. underlying loss episode frequency and duration The experiments demonstrate the trade-offs between characteristics. impact on the network and measurement accuracy. We show that BADABING reports loss characteristics far Advantages: more accurately than traditional loss measurement tools. The most important implication of these results Obtainable System: We begin by testing the capability of is that there is now a methodology and tool standard Poisson-modulated end-to-end measurements of available for wide-area studies of packet loss loss in a controlled laboratory environment using IP characteristics that enables researchers to routers and commodity end hosts. Our tests show that understand and specify the trade-offs between loss characteristics reported from such Poisson- accuracy and impact. modulated probe tools can be quite inaccurate over a Furthermore, the tool is self-calibrating in the sense range of traffic conditions (10). that it can report when estimates are poor. measurement tool called BADABING. BADABING