A system for online compression of high-speed network measurements

2008-10-24T11:00:52Z (GMT) by Kostas Kyriakopoulos David Parish
Measuring various metrics of high speed and high capacity networks pro- duces a vast amount of information over a long period of time, making the conventional storage of the data practically ine cient. Such metrics are derived from packet level information and can be represented as time series signals. Thus, they can be ana- lyzed using signal analysis techniques. This paper looks at the Wavelet transform as a method of analyzing and compressing measurement signals (such as delay, utilization, data rate etc.) produced from high-speed networks. A live system can calculate these measurements and then perform wavelet techniques to keep the signi cant information and discard the small variations. An investigation into the choice of an appropriate wavelet is presented along with results both from o -line and on-line experiments. The quality of the decompressed signal is measured by the PSNR and a comparison of compression performance is presented against the lossless tool bzip2.