A Framework for the Busy Time Calculation of Multiple Correlated Events

Many approaches to determine the response time of a task have difficulty to model tasks with multiple memory or coprocessor accesses with variable access times during the execution. As the request times highly depend on system setup and state, they can not be trivially bounded. If they are bounded by a constant value, large discrepancies between average and worst case make the focus on single worst cases vulnerable to overestimation. We present a novel approach to include remote busy time in the execution time analysis of tasks. We determine the time for multiple requests by a task efficiently and and far less conservative than previous approaches. These requests may be disturbed by other events in the system. We show how to integrate such a multiple event busy time analysis to take into account behavior of tasks that voluntarily suspend themselves and require multiple data from remote parts of the system.