Online Machine Learning for Energy-Aware Multicore Real-Time Embedded Systems

Abstract : In this paper, we present an Online Learning Artificial Neural Network (ANN) model that is able to predict the performance of tasks in lower frequency levels and safely optimize real-time embedded systems' power saving operations. The proposed ANN model is supported by feature selection, which provides the most relevant variables to describe shared resource contention in the selected multicore architecture. The variables are used at runtime to produce a performance trace that encompasses sufficient information for the ANN model to predict the impact of a frequency change on the performance of tasks. A migration heuristic encompassing a weighted activity vector is combined with the ANN model to dynamically adjust frequencies and also to trigger task migrations among cores, enabling further optimization by solving resource contentions and balancing the load among cores. The proposed solution achieved energy-savings of 24.97% on average when compared to the run-to-end approach, and it did it without compromising the criticality of any single task. The overhead incurred in terms of execution time was 0.1791% on average. Each prediction added 15.3585us on average and each retraining cycle triggered at frequency adjustments was never larger than 100us.
 EXISTING SYSTEM :
 ? Existing policies can be broadly classified into predictive schemes and stochastic schemes according to authors in . ? Predictive policies predict the duration of upcoming idle period as soon as a system component goes idle and the decision to transition state can be made if the prediction indicates a long idle period. ? Stochastic policies take into account both power consumption as well as performance penalty. ? They take a set of existing DPM policies and design a control mechanism that selects, in an online fashion, the best-suited policy for a given idle period.
 DISADVANTAGE :
 ? The optimization techniques like Genetic, Particle Swarm, Stochastic Evolution, Ant, and others algorithms are used to determine the optimal solution of a problem. Recent researches in energy aware scheduling uses the optimization techniques to solve the Bin-packing and scheduling problem. ? Finally, the problem of partitioning tasks is considered as an NP- problem. So it is transformed into the binpacking problem or optimization techniques. Many heuristics are based on the variable sized bin packing techniques for multi-core system. ? Others are proved to yield better results when using optimization techniques such as, Ant Colony Optimization, Genetic, Particle Swarm, and others.
 PROPOSED SYSTEM :
 • we propose a generic power management scheme, called the Hybrid Power Management (HyPowMan) scheme. • Our proposed scheme takes a set of well-known existing policies, each of which performs well for a given set of conditions, and adapts at runtime to the best-performing technique for any given workload. • We demonstrate in this paper that our proposed scheme quickly converges to the best-performing technique at any point in time. • HyPowMan scheme enhances the ability of portable embedded systems to adapt with changing workload and give an overall performance and energy savings that is better than any single policy can offer. Rest of this paper is organized as follows.
 ADVANTAGE :
 • Two widely used techniques for energy aware real time scheduling are DVFS (Dynamic voltage Frequency Scaling) and DPM (Dynamic Power Management). • The power dissipation for processor is divided into static power and dynamic power. DVFS deals with enhancing the performance of system by decreasing the supply voltage leading to decrease the dynamic power and enhances the overall energy dissipation. • Taking into account that the decrease in frequency leads to increase in the task's execution time, so the task's time constrains must be taken into consideration . • DPM is used to determine a specific point in which the processor is switched into sleep mode to decrease the leakage current , leading to decrease in its static power

We have more than 145000 Documents , PPT and Research Papers

Have a question ?

Mail us : info@nibode.com