Simulating execution time and power consumption of real-time tasks on embedded platforms

Abstract

In this paper, we present PARTSim, an open-source power/thermalaware simulator for embedded real-time systems. This tool is a fork of the well-known RTSim simulator, which can simulate the timing behavior of a set of real-time tasks with various characteristics when running on a multi-processor platform in presence of a number of real-time scheduling policies. PARTSim extends the functionality of RTSim by introducing support for power-aware embedded platforms exhibiting frequency scaling and specific architectural patterns like the ARM big.LITTLE and DynamIQ ones. Experimental results that compare simulated data against execution profiles collected on real platforms show a simulation error under 10 % for both execution time and power consumption at 90th percentile when simulating the effects of DVFS. CCS CONCEPTS• Computer systems organization ✙ Embedded systems; Realtime system architecture; • Software and its engineering ✙Scheduling; Power management; • Hardware ✙ Temperature simulation and estimation; Power estimation and optimization;

Multi-criteria Optimization of Real-time DAGs on Heterogeneous Platforms under P-EDF

“…The placement of the tasks in the {𝑥 𝑖,𝑝 } variables and the choice of the operating frequencies of the islands in the {𝑦 𝑠,𝑚 } ones, affect the average power consumption of the applications deployed on the platform, which the solver attempts to minimize. Precisely, following the power model discussed in [5], based on extensive experimental power consumption data gathered for a set of diverse workloads on various embedded platforms, the objective of the…”

“…Each application has been run on both big and LITTLE cores, and for all of the available CPU frequencies. The data-set has been gathered using PARTProf, and embedded within the PARTSim open-source real-time systems simulator [5].…”

Development of Energy- Efficient Load Balancing Scheduling Algorithm in Real-time Systems

“…A simulation tool (SIMSO) presented in [5] is designed for the understanding of realtime scheduling algorithms and evaluate real-time system performance and overheads. SIMSO used by many studies such as [3], [14], [4] and [7].…”

“…This event can be the release of a new task or termination of a specific task or any other event related to the scheduling algorithm type. Figure (3) shows the difference between event driven and non-event driven scheduling in terms of response time.…”

“…∀ 𝑆 ∈ 𝑄 ⇒ 𝑡(𝑛) 𝑆 ≠ 𝑡(𝑛 + 1) 𝑆 (3)  Execution time (𝐸 𝑠 ): It is the processing time that occupied by the task being executed on the processor.…”