*** Entwicklung eines skalierbaren Data-Mining-basierten Vorhersagemodells für IKT- und Stromversorgungssysteme (ScaMPo) ***
The Institute for Automation of Complex Power Systems and the IT Center are concerned with problems that arise with ever-increasing complexity of power grids and supercomputers in the project "Development of a Scalable Data-Mining-Based Prediction Model for ICT and Power Systems" (ScaMPo). Changes or failures of components can affect performance as well as power consumption. The huge amount of data that is generated during the monitoring of such systems is usually too large to allow them to be surveyed for suitable interrelationships between the data. As a result, it is difficult to predict how certain changes will affect the networks or where failures are likely to occur.
ScaMPo is intended to provide a framework that can be used for scalable storage of data in a cloud environment as well as for efficient analysis and pattern recognition in these data. This framework should allow to generate predictions using suitable models and to derive suggestions for changes from them. This approach could be used to detect hardware failures or to reduce energy costs through appropriate measures.