Abhilash Thekkilakattil

The Global Limited Preemptive Earliest Deadline First Feasibility of Sporadic Real-Time Tasks

The feasibility of preemptive and non-preemptive scheduling has been well investigated on uniprocessor and multiprocessor platforms under both Fixed Priority Scheduling (FPS) and Earliest Deadline First (EDF) paradigms. While feasibility of limited preemptive scheduling under FPS has been addressed on both uniprocssor and multiprocessor platforms, under EDF it has been investigated only on uniprocessors, and a similar analysis for multiprocessor platforms is still missing. In this paper, we introduce global Limited Preemptive Earliest Deadline First (g-LP-EDF) scheduling, and propose the associated feasibility analysis to complete the above described feasibility analysis spectrum. Specifically, we derive a sufficient condition that guarantees g-LP-EDF feasibility of sporadic real-time tasks which directly provides a global Non-Preemptive Earliest Deadline First (g-NP-EDF) feasibility test. We then study the interplay between g-LP-EDF feasibility and processor speed, in order to quantify the sub-optimality of g-NP-EDF in terms of the minimum speed-up required to guarantee g-NP-EDF feasibility of all feasible task sets. The results presented in this paper complement our previous results on uniprocessors, and provide a unified result on the sub-optimality of non-preemptive EDF on both uniprocessor and multiprocessor platforms.

The limited-preemptive feasibility of real-time tasks on uniprocessors

The preemptive scheduling paradigm is known to strictly dominate the non-preemptive scheduling paradigm with respect to feasibility. On the other hand, preemptively scheduling real-time tasks on uniprocessors, unlike non-preemptive scheduling, may lead to unschedulability due to, e.g., preemption related overheads. The limited-preemptive scheduling paradigm, which is a generalization of preemptive and non-preemptive paradigms, has, however, the potential to reduce the preemption related overheads while enabling high processor utilization. In this paper, we focus on the characterization of the effects of increasing the computational resources on the limited-preemptive feasibility of real-time tasks in order to quantify the sub-optimality of limited-preemptive scheduling. Specifically, we first derive the required processor speed-up bound that guarantees limited-preemptive feasibility of any uniprocessor feasible taskset. Secondly, we demonstrate the applicability of the results in the context of controlling preemption related overheads while minimizing the required processor speed-up. In particular, we identify the preemptive behavior that minimizes preemption-related overheads, as well as derive the optimal processor speed associated with it. Finally, we examine the consequences of having more processors on limited-preemptive feasibility and derive the bound on the number of processors that guarantees a specified limited-preemptive behavior for any uniprocessor feasible real-time taskset. This paper essentially bridges the preemptive and non-preemptive real-time scheduling paradigms by providing significant theoretical results building on the limited-preemptive scheduling paradigm, as well as provides analytical inputs to developers in order to perform various trade-offs, e.g., code refactoring, to control the preemptive behavior of real-time tasks.

Quantifying the Exact Sub-optimality of Non-preemptive Scheduling

Fixed priority scheduling is used in many real-time systems, however, both preemptive and non-preemptive variants (FP-P and FP-NP) are known to be sub-optimal when compared to an optimal uniprocessor scheduling algorithm such as preemptive Earliest Deadline First (EDF-P). In this paper, we investigate the sub-optimality of fixed priority non-preemptive scheduling. Specifically, we derive the exact processor speed-up factor required to guarantee the feasibility under FP-NP (i.e. schedulablability assuming an optimal priority assignment) of any task set that is feasible under EDF-P. As a consequence of this work, we also derive a lower bound on the sub-optimality of non-preemptive EDF (EDF-NP), which since it matches a recently published upper bound gives the exact sub-optimality for EDF-NP. It is known that neither preemptive, nor non-preemptive fixed priority scheduling dominates the other, i.e., there are task sets that are feasible on a processor of unit speed under FP-P that are not feasible under FP-NP and vice-versa. Hence comparing these two algorithms, there are non-trivial speedup factors in both directions. We derive the exact speed-up factor required to guarantee the FP-NP feasibility of any FP-P feasible task set. Further, we derive upper and lower bounds on the speed-up factor required to guarantee FP-P feasibility of any FP-NP feasible task set. Empirical evidence suggests that the lower bound may be tight, and hence equate to the exact speed-up factor in this case.

Optimizing Preemption-Overhead Accounting in Multiprocessor Real-Time Systems

There exist two general techniques to account for preemption-related overheads on multiprocessors. This paper presents a new preemption-related overhead-accounting technique, called analytical redistribution of preemption overheads (ARPO), which integrates the two previous techniques to minimize preemption-overhead-related utilization loss. ARPO is applicable under any job-level fixed priority (JLFP) preemptive scheduler, as well as some limited-preemption schedulers. ARPO is evaluated in a new experimental-design framework for overhead-aware schedulability studies that addresses unrealistic simplifying assumptions made in previous studies, and is shown to improve real-time schedulability.

Probabilistic preemption control using frequency scaling for sporadic real-time tasks

Preemption related costs are major sources of unpredictability in the task execution times in a real-time system. We examine the possibility of using CPU frequency scaling to control the preemption behavior of real-time sporadic tasks scheduled using a preemptive Fixed Priority Scheduling (FPS) policy. Our combined offline-online method provides probabilistic preemption control guarantees by making use of the release time probabilities of the sporadic tasks. The offline phase derives the probability related deviation from the minimum inter-arrival time of tasks. The online algorithm uses this information to calculate appropriate CPU frequencies that guarantees non-preemptive task executions while preserving the overall system schedulability. The online algorithm has a linear complexity and does not lead to significant implementation overheads. Our evaluations demonstrate the effectiveness of the method as well as the possibility of energy-preemption trade offs. Even though we have considered FPS, our method can easily be extended to dynamic priority scheduling schemes.

Bounding the effectiveness of temporal redundancy in fault-tolerant real-time scheduling under error bursts

Reliability is a key requirement in many distributed real-time systems deployed in safety and mission critical applications, and temporal redundancy is a widely employed strategy towards guaranteeing it. The temporal redundancy approach is typically based on task re-executions in form of entire tasks, task alternates or, check-pointing blocks, and each of the re-execution strategies have different impacts on the Fault Tolerance feasibility (FT-feasibility) of the system, which is traditionally defined as the existence of a schedule that guarantees timeliness of all tasks under a specified fault hypothesis. In this paper, we propose the use of resource augmentation to quantify the FT-feasibility of real-time task sets and use it to derive limits on the effectiveness of temporal redundancy in fault-tolerant real-time scheduling under error bursts of bounded lengths. We derive the limits for the general case, and then show that for the specific case when the error burst length is no longer than half the shortest deadline, the lower limit on the effectiveness of temporal redundancy is given by the resource augmentation bound 2, while, the corresponding upper-limit is 6. Our proposed approach empowers a system designer to quantify the effectiveness of a particular implementation of temporal redundancy.

An Empirical Investigation of Eager and Lazy Preemption Approaches in Global Limited Preemptive Scheduling

Global limited preemptive real-time scheduling in multiprocessor systems using Fixed Preemption Points FPP brings in an additional challenge with respect to the choice of the task to be preempted in order to maximize schedulability. Two principal choices with respect to the preemption approach exist 1 the scheduler waits for the lowest priority job to become preemptible, 2 the scheduler preempts the first job, among the lower priority ones, that becomes preemptible. We refer to the former as the Lazy Preemption Approach LPA and the latter as the Eager Preemption Approach EPA. Each of these choice has a different effect on the actual number of preemptions in the schedule, that in turn determine the runtime overheads. In this paper, we perform an empirical comparison of the run-time preemptive behavior of Global Preemptive Scheduling and Global Limited Preemptive Scheduling with EPA and LPA, under both Earliest Deadline First EDF and Fixed Priority Scheduling FPS paradigms. Our experiments reveal interesting observations some of which are counter-intuitive. We then analyse the counter-intuitive observations and identify the associated reasons. The observations presented facilitate the choice of appropriate strategies when using limited preemptive schedulers on multiprocessor systems.

Towards preemption control using CPU frequency scaling in sporadic task systems

Preemptions in real-time systems scheduling typically lead to variations in task execution times, increase the temporal overhead required for various RTOS related operations and may even cause unschedulability. We examine the preemption behavior of sporadic tasks scheduled under the Fixed Priority Scheduling (FPS) policy, and evaluate the possibility of using CPU frequency scaling for preemption control. We propose an online heuristic-based algorithm, of linear complexity, to control the number of preemptions in a sporadic task system using CPU frequency scaling. Evaluation results show that CPU frequency scaling is an attractive option to control the preemption behavior of real-time sporadic task systems.

Exact speedup factors and sub-optimality for non-preemptive scheduling

Fixed priority scheduling is used in many real-time systems; however, both preemptive and non-preemptive variants (FP-P and FP-NP) are known to be sub-optimal when compared to an optimal uniprocessor scheduling algorithm such as preemptive earliest deadline first (EDF-P). In this paper, we investigate the sub-optimality of fixed priority non-preemptive scheduling. Specifically, we derive the exact processor speed-up factor required to guarantee the feasibility under FP-NP (i.e. schedulability assuming an optimal priority assignment) of any task set that is feasible under EDF-P. As a consequence of this work, we also derive a lower bound on the sub-optimality of non-preemptive EDF (EDF-NP). As this lower bound matches a recently published upper bound for the same quantity, it closes the exact sub-optimality for EDF-NP. It is known that neither preemptive, nor non-preemptive fixed priority scheduling dominates the other, in other words, there are task sets that are feasible on a processor of unit speed under FP-P that are not feasible under FP-NP and vice-versa. Hence comparing these two algorithms, there are non-trivial speedup factors in both directions. We derive the exact speed-up factor required to guarantee the FP-NP feasibility of any FP-P feasible task set. Further, we derive the exact speed-up factor required to guarantee FP-P feasibility of any constrained-deadline FP-NP feasible task set.

Preemption Control Using Frequency Scaling in Fixed Priority Scheduling

Controlling the number of preemptions in real time systems is highly desirable in order to achieve an efficient system design in multiple contexts. For example, the delays due to context switches account for high preemption overheads which detrimentally impact the system schedulability. Preemption control can also be potentially used for the efficient control of critical section behaviors in multi-threaded applications. At the same time, modern processor architectures provide for the ability to selectively choose operating frequencies, primarily targeting energy efficiency as well as system performance. In this paper, we propose the use of CPU Frequency Scaling for controlling the preemptive behavior of real-time tasks. We present a framework for selectively eliminating preemptions, that does not require modifications to the task attributes or to the underlying scheduler. We evaluate the proposed approach by four different heuristics through extensive simulation studies.