New algorithms for performance trace analysis based on compressed complete call graphs

Abstract

This paper addresses performance and scalability issues of state-of-the-art trace analysis. The Complete Call Graph (CCG) data structure is proposed as an alternative to the common linear storage schemes. By transparent in-memory compression CCGs are capable of exploiting redundancy as frequently found in traces and thus reduce the memory requirements notably. Evaluation algorithms can be designed to take advantage of CCGs, too, such that the computational effort is reduced in the same order of magnitude as the memory requirements. © Springer-Verlag Berlin Heidelberg 2005.