Combining Structured and Unstructured Data for Web Application Performance Analysis

System performance analysis is typically conducted by collecting structured data in the form of metrics or by examining unstructured data in the form of logs. Tools currently available either analyze the structured data or the unstructured data, but not both. In some settings, this is not enough to make meaningful statements about the performance of the system. In this work, we explore alternatives for querying system performance in a web application environment. PHP applications execute under the same process as the web server itself. This makes characterizing the performance of PHP applications more difficult because tools like UNIX Top only show process-level metrics. However, logs generated by the web server can also contain valuable information about the the web applications. Because these logs are sources of largely unstructured information, we use information retrieval techniques to query them. By combining the unstructured data from the logs with structured metrics from Top, we can assign resource usage estimates to the web applications.