Title: Applying Data Mining Techniques to Computer Systems

Abstract:
As computer systems are becoming increasingly complex, it is challenging to provide high performance, reliability and manageability that are demanded by enterprise customers. In order to provide a solution, it is necessary to first analyze and characterize such systems and identify the critical issues. The system data such as source code, running traces, etc. provide a valuable asset for us to target the solution. In the meantime, the huge amount of system data, however, renders a tedious and difficult task on managers and developers, and hence the hidden information would be difficult to extract for system characterization.

In this talk, I will present a novel approach to analyze various system data by applying data mining techniques. This approach can effectively obtain useful information hidden in huge amount of system data and then we can exploit such information to improve system performance, reliability and manageability. Specifically, we apply different data mining algorithms on different types of system data such as access traces, system call traces, and source code, to achieve different goals including automated debugging and system behavior characterization. The results demonstrate that using data mining techniques is an efficient and effective approach to solving computer system problems.

Bio:
Zhenmin Li is a PhD student supervised by Prof. Yuanyuan Zhou. His research interests include computer systems, software reliability, data mining, storage systems, and energy management. He received his BE and ME in computer science from Tsinghua University, China. Further information is available at http://opera.cs.uiuc.edu/~zli4.