Enhanced execution trace abstraction approach using social network analysis methods

Abstract

In this paper, we propose an improvement in system execution tracing by applying social network analysis techniques on the trace data. We perform a 3-step analysis: collection of trace data on operating system kernel; community analysis on the data; and PageRank algorithm within each community. The proposed analysis focused on the following problems: useless information contained in the data and the enormous size of the data. We propose two use cases: one on kernel trace filtering and the other on virtual machine clustering. Our evaluation shows that the proposed method provided a concise and more comprehensive view of the trace data. This can help shorten the time and assist in building infrastructural functions in analyzing system execution.