Online Real-Time Preemptive Scheduling of Jobs with Deadlines

IEEE Embedded Systems Letters, 2015

In this paper, processor speedup analysis is used to strengthen recent results regarding the sub-optimality of uniprocessor non-preemptive Earliest Deadline First (npEDF) scheduling. The sub-optimality of npEDF is defined as the minimum amount of increase in the processor speed that is needed to guarantee the npEDF schedulability of any feasible task set. We show that any preemptively schedulable task set that is not schedulable by npEDF will become schedulable on a processor speeded up by a factor of not more than one plus the value of the largest execution requirement divided by the shortest relative deadline of any task. This reduces the pessimism compared to the best previous bound by factor of at least two. In addition, for the case of non-preemptive Fixed Priority scheduling, we also show that twice this speedup bound is enough to guarantee the schedulability of any feasible task set.

2012

A number of preemptive real-time scheduling algorithms, such as Earliest Deadline First (EDF), are known to be optimal on uni-processor systems under specified assumptions. However, no uni-processor optimal algorithm exists under the non-preemptive scheduling paradigm. Hence preemptive schemes strictly dominate non-preemptive schemes with respect to uni-processor feasibility. However, the 'goodness' of non-preemptive schemes, compared to uni-processor optimal preemptive scheduling schemes such as EDF, which can also be referred to as its sub-optimality, has not been fully investigated yet. In this paper, we apply resource augmentation, specifically processor speed-up, to quantify the sub-optimality of non-preemptive scheduling with respect to EDF, and apply the results to guarantee user specified upper-bounds on the preemption related scheduling costs. In particular, we derive an upper bound on the minimum processor speed-up required to guarantee the non-preemptive feasibility of tasks that are deemed feasible under the preemptive EDF, and we prove that, in the cases where, for all tasks in the task set, the largest execution requirement is not greater than the shortest deadline, this bound is equal to 4. We also show how the proposed approach enables a system designer to choose an optimal processor, in order to, e.g., guarantee specified upperbounds on the preemption related overheads.

Proceedings of the 30th on Symposium on Parallelism in Algorithms and Architectures - SPAA '18, 2018

When a computer system schedules jobs there is typically a significant cost associated with preempting a job during execution. This cost can be from the expensive task of saving the memory's state and loading data into and out of memory. It is desirable to schedule jobs non-preemptively to avoid the costs of preemption. There is a need for non-preemptive system schedulers on desktops, servers and data centers. Despite this need, there is a gap between theory and practice. Indeed, few non-preemptive online schedulers are known to have strong foundational guarantees. This gap is likely due to strong lower bounds on any online algorithm for popular objectives. Indeed, typical worst case analysis approaches, and even resource augmented approaches such as speed augmentation, result in all algorithms having poor performance guarantees. This paper considers on-line non-preemptive scheduling problems in the worst-case rejection model where the algorithm is allowed to reject a small fraction of jobs. By rejecting only a few jobs, this paper shows that the strong lower bounds can be circumvented. This approach can be used to discover algorithmic scheduling policies with desirable worst-case guarantees. Specifically, the paper presents algorithms for the following two objectives: minimizing the total flow-time and minimizing the total weighted flow-time plus energy under the speed-scaling mechanism. The algorithms have a small constant competitive ratio while rejecting only a constant fraction of jobs. Beyond specific results, the paper asserts that alternative models beyond speed augmentation should be explored to aid in the discovery of good schedulers in the face of the requirement of being online and non-preemptive.

European Journal of Operational Research, 2009

We consider a single resource that performs a set of jobs, each job being defined by its operation time and its due date. Jobs are not preemptive, which means that a job that starts cannot be interrupted before completion. These jobs have been scheduled so as to minimize the total tardiness, and the schedule is known.

2020 IEEE 26th International Conference on Embedded and Real-Time Computing Systems and Applications (RTCSA), 2020

Different global and semi-partitioned schedulers have been proposed that are soft-real-time (SRT) optimal for sporadic task systems, meaning they can guarantee bounded deadline tardiness. However, under known analyses, tardiness bounds increase with respect to the number of processors, which reduces the applicability of these schedulers in systems with a large number of processors. In this paper, a semi-clustered scheduler, SC-EDF, is presented that has a constant tardiness bound. SC-EDF partitions tasks into clusters, each of which may include one fractional processor. Each cluster is scheduled by G-EDF, and the fractional processors are realized using Pfair scheduling techniques.

2015

Real-time systems are traditionally classified into hard real-time and soft real-time: in the first category we have safety critical real-time systems where missing a deadline can have catastrophic consequences, whereas in the second class we find systems or which we need to optimise the Quality of service provided to the user. However, the frontier between these two classes is thinner than one may think, and many systems that were considered as hard real-time in the past should now be reconsidered under a different light. In this paper we shall first recall the fundamental notion of time-predictability and criticality, in order to understand where the real-time deadlines that we use in our theoretical models come from. We shall then introduce the model of a soft real-time system and present one popular method for scheduling hard and soft real-time tasks, the resource reservation framework. Finally, we shall show how resource reservation techniques can be successfully applied to the...

Operations Research Letters, 2000

This note deals with the scheduling problem of maximizing the number of early jobs on a single machine. We investigate the on-line version of this problem in the Preemption-Restart model. This means that jobs may be preempted, but preempting results in all the work done on this job so far being lost. Thus, if the job is restarted, then it has to be done from scratch.

Computing Research Repository, 2011

An optimal solution to the problem of scheduling real-time tasks on a set of identical processors is derived. The described approach is based on solving an equivalent uniprocessor real-time scheduling problem. Although there are other scheduling algorithms that achieve optimality, they usually impose prohibitive preemption costs. Unlike these algorithms, it is observed through simulation that the proposed approach produces no more than three preemptions points per job.

In this paper we consider the online scheduling of jobs, which require processing on a number of machines simultaneously. These jobs are presented to a decision maker one by one, where the next job becomes known as soon as the current job is scheduled. The objective is to minimize the makespan. For the problem with three machines we give a

As Network Of Workstations (NOWs) emerge as a viable platform for a wide range of workloads, new scheduling approaches are needed to allocate the collection of resources from competing applications. New workload types introduce high uncertainty into the predictability of the system, hindering the applicability of the job scheduling strategies. A new kind of parallel applications has appeared in business or scientific domains, namely Soft Real-Time (SRT). They, together with new SRT desktop applications, turn prediction into a more difficult goal by adding inherent complexity to estimation procedures. In previous work, we introduced an estimation engine into our job scheduling system, termed CISNE. In this work, the estimation engine is extended, by adding two new kernels, both SRT aware. Experimental results confirm the better performance of simulated respect to the analytical kernels and show a maximum average prediction error deviation of 20%.