In this paper, we develop an online monitoring methodology for the time-varying cyclical streams of network data, that combines a baseline state-space model and statistical control schemes to monitor departures from the baseline model. Parameters of the state space models are initialized based on the training data, and updated for each incoming observation. The statistical control schemes for monitoring are designed based on forecasting errors from the baseline model, under the framework of statistical change detection. We demonstrate the effectiveness of our methodology using measurement data from an EVDO wireless network.

We have performed extensive measurements upon commercial cdma2000 networks supporting predominantly voice services using a proprietary monitoring tool called Celnet Xplorer. These measurements capture large numbers of user-mobile calls under actual operating conditions. We also capture statistically significant number of user-mobile call failures, in particular lost calls. This massive data set allows us to correlate call failures against the network operating conditions at the same moment in time, in order to infer underlying cause and effect relationships. We observe that the call failure probability has a strong exponential dependence upon the total forward-link power on the dominant interfering cell-sector. We also conclude that the dominant physical source of this increase in lost call rate is a loss of coverage at increased load, and not loading-induced diversity reduction. We note that these results support an important new paradigm where network analyses are based upon real-time measurements of actual user mobiles