The System Energy Minimization for Weakly Hard Real Time System Using (m,k) Variables

Abstract

Energy consumption and quality of service (QoS) are the two primary concerns in the development of today’s pervasive computing systems. While most of the current research in energy-aware real-time scheduling has been focused on Hard Real-time Systems, a large number of practical applications and systems exhibit more weakly hard real-time. Weakly Hard Real-time Systems can tolerate some occasional deadline misses. This feature provides a unique opportunity to reduce system’s energy.
Our goal is to minimize the system energy (energy required by frequency dependent and independent component) rather than minimization of processor energy only. We use the term frequency dependent component to refer a processor and frequency independent for memory or peripheral devices. We aim to minimize the system energy for weakly hard real time systems modeled with constraint using a combination of Dynamic voltage scaling (DVS) and Dyanamic power down (DPD). The QoS requirements are deterministically quantified with imprecise concept or by (m, k) model, while energy minimization is done in two phases, in the first phase the feasibility and energy reduction at the task level is achieved while further reduction in the energy consumption is accomplished in the improvement (second) phase at the job level. We propose a new portioning strategy to decide a job to be mandatory or optional with speed assignment for each task is done based on the greedy speed assignment technique in phase 1. While in the second phase we adopt the preemption control technique by delaying the higher priority jobs without missing its deadline. Experiments were performed and it was found that our proposed techniques improve significantly in terms of both the energy minimization as well as QoS over the existing one and over wide range of parameter including variation of m & k.