Certain investigations on energy efficient real time embedded system using power management techniques

Abstract

In Real-Time Embedded System (RTES), energy consumption and reliability are two major design constraints. More consumption of energy results in lower battery life and greater temperatures. In particular, the issue of energy management in embedded systems is worse as most of these systems have restricted power supplies. RTES, typically use system level energy management methods, i.e., Dynamic Voltage Scaling (DVS) and Dynamic Power Management (DPM). DVS is recognized to be effective in reducing processor energy consumption, which has received a lot of attention in recent years. DPM is a general power management method that can be implemented to any embedded system element. DVS saves energy by running the system at a reduced frequency and thus a reduced voltage, while DPM saves energy by placing the system in a reduced power state when the idle time is long enough. Hard Real-Time Embedded System (HRTES) also frequently employs standby-sparing system consisting of two identical processors. One of them is called the main unit, while the other is called the spare unit, replacing the main unit when it becomes faulty. The impact of system level energy reduction methods on reliability and energy consumption of HRTES is investigated. There are two existing energy management methods are considered: 1) Postponement method: a variation of DPM which is only applicable to replicated systems, and 2) Hybrid method: a combination of Postponement and DVS techniques. In these existing methods, when reliability is the primary concern, the postponement method is the most preferable. However, the hybrid method is more suitable if the energy consumption is the primary concern. Due to the increasing importance of the energy consumption of the memory subsystem, Dynamic Cache Reconfiguration (DCR) technique is recently proposed to save the energy consumption of the cache subsystem. DCR is very effective to reduce energy consumption of cache subsystems which accounts for about half of the total energy consumption in embedded systems.