timings, this might not be always the case. We suggest therefore a modification of
the above algorithm to take this into account.
The main idea is to include a task ready time t ri for every task in the system. The
task finishing time changes accordingly to t i = t ri + t pi , i.e., the ready time (t ri ) + the
task processing time t pi of task number i.
In addition, we want to extend the task checking from the processor only to all
the hardware that is used by the task and introduce t adi , the asynchronous task
checking time for task number i.
Taking this change into account, we modify the procedure in Fig. 7.6.
First, we want to keep the real-time nature of the algorithm and sort the tasks
according to their finishing time (earliest deadline first). Except for the continuously
running tasks, which can be tested at any time, it is advisable to move the testing to
the end of a task processing to minimize recovery time in case of a fault. The
modified algorithm is shown in Fig. 7.8.
Fig. 7.8 Procedure T2
82
7 Testing, Checking, and Hardware Syndrome