Skip to Main Content
This paper presents a toolset formed by a timing library and a clock synchronization library for a cluster. These are tools developed to measure both execution and communication times in distributed environments for performance evaluation. The toolset scalability is shown by experimentation as well as the instrumentation of a parallel matrix multiplication algorithm, based on broadcast communications. The performance evaluation toolset overhead is analyzed per depending on the number of computers. Possible performance penalization is identified by the accurate data provided by the toolset.