Selective monitoring using performance metric predicates

Abstract

The field of parallel processing is going through an important evolution in technology characterized by a significant increase in the number of processors within such systems. As the number of processors increases, the conventional techniques for monitoring the performance of parallel systems will produce large amounts of data in the form of event trace files. The authors propose one possible solution to this data size problem: performance metric predicates. These predicates permit the user to define performance parameters that control the output of event trace data during the application's execution time. The authors assert that the use of performance metric predicates provides a powerful and useful tool for the control of event trace data output from large, complex systems.<>