Power Management in Embedded Systems with Hard and Soft Real-time Tasks

Please enable javascript in your browser.

Toutounchian, Pouya | 2017

411 Viewed

Type of Document: M.Sc. Thesis
Language: Farsi
Document No: 49999 (19)
University: Sharif University of Technology
Department: Computer Engineering
Advisor(s): Ejlali, Alireza
Abstract:
Embedded systems usually have limited energy budget, therefore energy management for these systems is important. On the other hand, many embedded systems have real-time and reliability constraints and such constraints are usually serious obstacles to effective energy management. Considering these tradeoffs between energy consumption and system reliability and timeliness, in this thesis, we consider system-level energy management techniques for embedded systems with both hard real-time and soft real-time tasks. Due to large search space of the problem, we use Genetic Algorithm in our proposed technique. Our proposed energy management technique guarantees the deadlines of hard real-time tasks and also attempts to increase the utility of soft real-time tasks. Experimental results show that the proposed technique decreased total energy consumption by 19.6% and dynamic energy consumption by 30.7% for a multicore system with four processing elements. This method, in the worst case, degraded utility accrual by 15%
Keywords:
Power Management ; Embedded System ; Genetic Algorithm ; Hard Real Time Systems ; Soft Real-Time Embedded ; Power Consumption