Podrobný výpis o publikaci

Time Series Analyzer is a customized tool for high-speed analysis of time series of network performance metrics. Being programmed in C++, the tool is able to process over thousand of these time-series in parallel providing alerts and statistical characteristics of the time series to an ITOA operator. The tool is customized for following use-cases: trend prediction and baseline modeling. The trend prediction analyses data from a given time-window, identifies the trend in the time series, and predict the value of the trend for the predefined interval. The baseline analysis creates a model of the typical behavior of the given time series and reports the observation that does not fit the model. The created model is adaptive and learns over time. Time series analyzer is designed to be connected over a network to a messaging system that provides it with time series data and receives the computed statistics and alerts.