Correlation analysis techniques for refining execution time estimates of real-time applications

Abstract

Scheduling techniques based upon worst case execution times, as are commonly used in real-time applications, often result in severe underutilization of the processor resources since most tasks finish in much less time than their anticipated worst-case execution times. We describe techniques for identifying correlation among the executions of various statements within a program. We demonstrate how this information can be used to refine the estimate of remaining worst case execution time of a real-time task as the execution of the task progresses. Refined estimates can be used at run-time to achieve better utilization, of the system and early failure detection and recovery.<>