... this paper we propose a solution to this problem; to our knowledge, this is the first solution that combines the three constraints mentioned above. We devise two algorithms, an optimal algorithm for homogeneous applications (with respect to power consumption) and a heuristic iterative algorithm that can also accommodate heterogeneous applications (that is, those with different power consumption functions). We show by simulation that our iterative algorithm is fast and within 1% of the optimal.

In the context of battery-powered real-time systems three constraints need to be addressed: energy, deadlines and task rewards. Many future real-time systems will count on different software versions, each with different rewards, time and energy requirements, to achieve a variety of QoS-aware tradeoffs. We propose a solution that allows the device to run the most valuable task versions while still meeting all deadlines and without depleting the energy. Assuming that the battery is rechargeable, we also propose (a) a static solution that maximizes the system value assuming a worstcase scenario (i.e., worst-case task execution times); and (b) a dynamic scheme that takes advantage of the extra energy in the system when worst-case scenarios do not happen. Three dynamic policies are shown to make better use of the recharging energy while improving the system value. 1

Wireless and portable devices depend on the limited power supplied by the battery. Dynamic Voltage Scaling (DVS) is an effective method to reduce CPU power consumption by adjusting CPU voltage. For real-time systems, DVS algorithms must also guarantee that no job misses its deadline. In this paper, we propose an integrated approach for applying DVS to real-time systems. We define two functions, the available cycle function (ACF) and the required cycle function (RCF), to capture the CPU workload of the real-time tasks under a scheduling policy. We then formulate the DVS scheduling problem for real-time systems as a nonlinear optimization problem and propose an optimal off-line algorithm to solve this problem. We also propose a novel online algorithm with time complexityÇ to further reduce power consumption when a job uses fewer execution cycles than the worst-case budget. The algorithms in this paper are based solely on ACF and RCF. When ACF and RCF are defined, the algorithms can be applied to any scheduling policy. We illustrate the generality of our approach over previous research by applying our method to several scheduling policies, including FIFO, EDF and RM. Our simulation results show significant improvement over previous work. 1

We present an energy-efficient real-time scheduling algorithm called EUA∗, for the unimodal arbitrary arrival model (or UAM). UAM embodies a “stronger ” adversary than most arrival models. The algorithm considers application activities that are subject to time/utility function time constraints, UAM, and the multi-criteria scheduling objective of probabilistically satisfying utility lower bounds, and maximizing system-level energy efficiency. Since the scheduling problem is intractable, EUA ∗ allocates CPU cycles, scales clock frequency, and heuristically computes schedules using statistical estimates of cycle demands, in polynomial-time. We establish that EUA ∗ achieves optimal timeliness during under-loads, and identify the conditions under which timeliness assurances hold. Our simulation experiments illustrate EUA∗’s superiority. 1.

Procrastination scheduling has gained importance for energy efficiency due to the rapid increase in the leakage power consumption. Under procrastination scheduling, task executions are delayed to extend processor shutdown intervals, thereby reducing the idle energy consumption. We propose algorithms to compute the maximum procrastination intervals for tasks scheduled by either the fixed priority or the dual priority scheduling policy. We show that dual priority scheduling always guarantees longer shutdown intervals than fixed priority scheduling. We further combine procrastination scheduling with dynamic voltage scaling to minimize the total static and dynamic energy consumption of the system. Our simulation experiments show that the proposed algorithms can extend the sleep intervals up to 5 times while meeting the timing requirements. The results show up to 18% energy gains over dynamic voltage scaling.

Multiprocessor implementations of real-time systems tend to be more energy-ecient than uniprocessor implementations: since the power consumed by a CMOS processor is approximately proportional to the cube of the speed or computing capacity at which the processor executes, the total power consumed by an m-processor multiprocessor platform is approximately (1=m ) times the power consumed by a uniprocessor platform of the same computing capacity. However several factors, including the non-existence of optimal multiprocessor scheduling algorithms, combine to prevent all the computing capacity of a multiprocessor platform from being guaranteed available for executing the real-time workload. In this paper, this tradeo | that while increasing the number of processors results in lower energy consumption for a given computing capacity, the fraction of the capacity of a multiprocessor platform that is guaranteed available for executing real-time work decreases as the number of processors increases | is explored in detail.

Slowdown factors determine the extent of slowdown a computing system can experience based on functional and performance requirements. Dynamic Voltage Scaling (DVS) of a processor based on slowdown factors can lead to considerable energy savings. We address the problem of computing slowdown factors for dynamically scheduled tasks with specified deadlines. We present an algorithm to compute a near optimal constant slowdown factor based on the bisection method. As a further generalization, for the case of tasks with varying power characteristics, we present the computation of near optimal slowdown factors as a solution to convex optimization problem using the ellipsoid method. The algorithms are practically fast and have the same time complexity as the algorithms to compute the feasibility of a task set. Our simulation results show on an average 20% energy gains over known slowdown techniques using static slowdown factors and 40 % gains with dynamic slowdown. 1.

Introduction A computerized real-time system is required to complete its work on a timely basis. Typical applications are digital control, command and control, signal processing and communication systems. The aim of real-time scheduling is to build-up a sequence of jobs that meets hard timing constraints at run-time. In opposition with the classical scheduling literature, real-time scheduling is not necessarily on-line scheduling. The main characteristic of real-time systems is the behavioral predictability. Timing constraints will be met whatever happens in the system. Real-time operating systems define recurring tasks. Real-time tasks are often periodically released. Basically, a periodic task gets its input data from sensors and command the controlled process by sending data to actuators. Every execution of a task is called a job. In the general model of periodic tasks, a task # i is defined by an o#set O i , that defines the release time of its first job; its worst-case execution

This dissertation was presented by Phongsak Kiratiwintakorn It was defended on

In a wide variety of embedded control applications, it is often the energy resources that form the fundamental limits on the system, not the system's computing capacity. Various techniques have been developed to improve energy efficiency in hardware, such as Dynamic Voltage Scaling (DVS), effectively extending the battery life of these systems. However, a comprehensive mechanism of task adaptation is needed in order to make the best use of the available energy resources, even in the presence of DVS or other power-reducing mechanisms. Further complicating this are the strict timeliness guarantees required by real-time applications commonly found in embedded systems. This paper