In this work with colleagues at AT&T, we were among the first to investigate how web services impacted the performance of grid platforms. Unlike earlier approaches, we pioneered techniques that allowed us to map our raw measurements unto Globus' internal architecture: We aggregated stack trace samples according to structural information and stack depth, thus addressing the growing complexity of key middleware platforms. This work was published at HPDC, a rigorously peer-reviewed conference (acceptance rate: 18.0%), and has since raised interest in the industry: I was invited to IBM Hursley in May 2006 to discuss our results. RAE_import_type : Conference contribution RAE_uoa_type : Computer Science and Informatics