Home

The Modaclouds Statistical Data Analyser (SDAs) is an important component of the Modaclouds monitoring platform. It extracts hidden information from the data using statistical methods and generate predictions. An example of hidden information estimated by SDAs are the resource requirements of individual requests when only aggregate measurements are available.

The SDAs are classified in the following categories: Correlation SDAs, that calculate the correlation between monitoring data (e.g., between a web server arrival traffic and traffic consequently generated onto the database tier); Estimation SDAs, which estimate QoS metrics of the system that cannot be directly measured (e.g., resource consumption of individual requests corrected to ignore contention); Forecasting SDAs, which forecast the trends of selected metrics using statistical methods. Among them, the Correlation SDAs and the machine learning based Forecasting SDAs are provided by the Modaclouds-SDAWeka.

The other SDAs are contained in the MATLAB Compiler Runtime (MCR). The MCR is a royalty-free container and no MATLAB license is required to execute SDAs. With minor changes, most SDAs should be able to run also inside GNU Octave.

The monitoring data is obtained through the Modaclouds DataRetriever and the configurations are provided with the Modaclouds KnowledgeBase API.

These wiki pages may be relevant to users:

• Data Analysers: A detailed description of each data collector provided by Modaclouds-SDA.
• Configuration File: Explain the configuration file for each data analyser.
• Installation: The installation steps to use SDAs.
• Compilation: Describe the procedures of compiling the MATLAB scripts.
• Common data format: Describe the common data format used for the Estimation SDAs.