S. Punnekkat. "Schedulability Analysis for Fault Tolerant Real-Time Systems". PhD thesis, Department of Computer Science, University of York, 1997.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....This characteristic is very useful since faulty tasks certainly have a shorter period of time to meet their deadlines. Due to its flexibility and simplicity, the proposed approach provides an effective schedulability analysis, where system predictability can be fully guaranteed. a recent approach [3, 13] based on the well known response time analysis [1] has eliminated these drawbacks by not restricting the way that fault tolerance is carried out and by assuming realistic fault and task models. In this work we generalise this approach by allowing task recovery to execute with higher priority, ....

....each task set, where p 1. 10. Then we set TF to the maximum found T i. This procedure is impor tant since it keeps the same value of TF per task set throughout the simulation, allowing a more accurate comparison. The procedure to find T i was implemented as a binary search on equation (3) as [13] suggests. Initially, the interval of search is [raaxvi(Ci) maxvi(Di) This initial interval is appropriate because: i) TF cannot assume lower values since faults at this rate or higher will continually disrupt the recovery of the tasks; ii) any larger value for TF will imply only one fault ....

S. Punnekkat. Schedulability analysis for fault tol- erant real-time systems. DPhil thesis, Department of Computer Science, University of York, 1997.

....the work in [10] extends the Rate Monotonic Scheduling (RMS) scheme and its exact characterization to provide tolerance for single and multiple transient faults. An important development in this work is the derivation of a Fault Tolerant RMS schedulability utilization bound. The work developed in [19] provided a feasibility analysis, using the worst case response time test for several models of fault recovery, namely re execution, recovery blocks, forward recovery and checkpointing. In the first part of our work, we extend this work to include optional computations. However, since the ....

....analysis, using the worst case response time test for several models of fault recovery, namely re execution, recovery blocks, forward recovery and checkpointing. In the first part of our work, we extend this work to include optional computations. However, since the schedulability tests in [19, 10] provide only a yes no answer to the schedulability question for sets of tasks without optional parts, it is necessary to develop an analytical solution or a reasonable heuristic strategy for selecting which optional parts should be shed when the system becomes overloaded. The problem of ....

[Article contains additional citation context not shown here]

S. Punnekkat. "Schedulability Analysis for Fault Tolerant Real Time Systems", PhD. Thesis, Dept. Computer Science, University of York, June 1997.

....C F j delay will be suffered by task i , where C F j denotes the timing requirement of the recovery operation of task j . Therefore, recovery will add max fj=1; ig C F j to the response time of task i . Note that this is true if only one fault occurs, that is, T F Tn . Punnekkat [10] extended the worst case response time analysis [1] to include fault tolerant real time tasks, calculating the response time by, r w 1 i = C i X j=1; i Gamma1 d r w i T j eC j d r w i T F emax fj=1; ig C F j (4) where T F denotes a fixed minimum time between ....

S. Punnekkat. "Schedulability Analysis for Fault Tolerant Real Time Systems", PhD. Thesis, Dept. CS, University of York, June 1997.

....where a criticality driven transient server is developed. The transient server creates server tasks in response to transient recovery requests. Its implementation requires the creation of a table before run time to assist the scheduler in performing on line decisions during recovery. Punnekkat [17] extended the analysis in [12] to develop a worst case response time analysis for different fault tolerant mechanisms, and developed probabilistic guarantees for fault tolerant realtime tasks. In [5, 8] a recovery scheme is proposed to extend the RMS scheme for single and multiple transient ....

S. Punnekkat. "Schedulability Analysis for Fault Tolerant Real Time Systems", PhD. Thesis, Dept. Computer Science, University of York, June 1997.

....that the deadline of task 3 is reduced to 140. Theorem 1. The above algorithm CP is optimal in the sense that if a task set is schedulable by any arbitrary combination of checkpoints, then it is schedulable by algorithmCP as well. The proof of the above theorem is by induction and is given in [6]. 4. Optimum Number of Checkpoints We analyse the problem of finding optimal number of checkpoints in two stages. First we try to find the individual optimum number of checkpoints for each task. Having found the individual optimum for each task, then we explore their mutual dependencies. 4.1. ....

S. Punnekkat. Schedulability Analysis for Fault Tolerant Real-time Systems. PhD thesis, Dept. Computer Science, University of York, June 1997.

....will be affected by a fault in i or any higher priority task. We assume that any extra computation for a task will be executed at the task s (fixed) priority. Probabilistic Scheduling Guarantees for Fault Tolerant Real Time Systems Hence if there is just a single fault, equation (1) will become [16, 2] 4 : R i = C i B i X j2hp(i) R i T j C j max k2hep(i) F k (4) where hep(i) is the set of tasks with priority equal or higher than i , that is hep(i) hp(i) i . This equation can again be solved for R i by forming a recurrence relation. If all R i values are still less ....

....will become unschedulable when faced with an arbitrary number of fault events. To consider maximum arrival rates, first assume that T f is a known minimum arrival interval for fault events. Also assume the error latency is zero (this restriction will be removed shortly) Equation (4) becomes [16, 2]: R i = C i B i X j2hp(i) R i T j C j R i T f max k2hep(i) F k (5) Thus in interval (0 R i ] there can be at most l R i T f m fault events, each of which can induce F k amount of each computation. The validity of this equation comes from noting that fault events ....

S. Punnekkat. Schedulability Analysis for Fault Tolerant Real-time Systems. PhD thesis, Dept. Computer Science, University of York, June 1997.

....that the deadline of task 3 is reduced to 140. Theorem 1. The above algorithm CP is optimal in the sense that if a task set is schedulable by any arbitrary combination of checkpoints, then it is schedulable by algorithmCP as well. The proof of the above theorem is by induction and is given in [6]. 4. Optimum Number of Checkpoints We analyse the problem of finding optimal number of checkpoints in two stages. First we try to find the individual optimum number of checkpoints for each task. Having found the individual optimum for each task, then we explore their mutual dependencies. 4.1. ....

S. Punnekkat. Schedulability Analysis for Fault Tolerant Real-time Systems. PhD thesis, Dept. Computer Science, University of York, June 1997.

....analysis the execution of task i will be affected by a fault in i or any higher priority task. Note, we assume that as each task has a fixed priority the extra computation time needed will also be executed at this fixed priority. Hence if there is just a single fault, equation (1) will become [Punnekkat 1997, Burns, Davis, Punnekkat 1996] 3 : R i = C i B i X j2hp(i) R i T j C j max k2hep(i) F k (4) 3 We assume that in the absence of faults, the task set is schedulable. where hep(i) is the set of tasks with priority equal or higher than i , that is hep(i) hp(i) i . This ....

....will become unschedulable when faced with an arbitrary number of fault events. To consider maximum arrival rates, first assume that T f is a known minimum arrival interval for fault events. Also assume the error latency is zero (this restriction will be removed shortly) Equation (4) becomes [Punnekkat 1997, Burns, Davis, Punnekkat 1996] R i = C i B i X j2hp(i) R i T j C j R i T f max k2hep(i) F k (5) Thus in interval (0 R i ] there can be at most l R i T f m fault events, each of which can induce F k amount of each computation. The validity of this equation comes ....

S. Punnekkat. Schedulability Analysis for Fault Tolerant Real-time Systems. PhD Thesis, Dept. Computer Science, University of York, 1997.

No context found.

S. Punnekkat. "Schedulability Analysis for Fault Tolerant Real-Time Systems". PhD thesis, Department of Computer Science, University of York, 1997.

No context found.

S. Punnekkat. Schedulability Analysis for Fault Tolerant Real-time Systems. PhD thesis, Dept. Computer Science, University of York, 1997.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC