A Survey of Techniques For Improving Energy Efficiency in Embedded Computing Systems (1401.0765v1)

Abstract: Recent technological advances have greatly improved the performance and features of embedded systems. With the number of just mobile devices now reaching nearly equal to the population of earth, embedded systems have truly become ubiquitous. These trends, however, have also made the task of managing their power consumption extremely challenging. In recent years, several techniques have been proposed to address this issue. In this paper, we survey the techniques for managing power consumption of embedded systems. We discuss the need of power management and provide a classification of the techniques on several important parameters to highlight their similarities and differences. This paper is intended to help the researchers and application-developers in gaining insights into the working of power management techniques and designing even more efficient high-performance embedded systems of tomorrow.

Energy Efficiency in Embedded Computing Systems: A Comprehensive Survey

The paper "A Survey of Techniques For Improving Energy Efficiency in Embedded Computing Systems" by Sparsh Mittal provides an extensive examination of power management strategies for embedded systems. It addresses core challenges and advances in managing energy consumption in increasingly ubiquitous embedded systems, such as mobile devices, which are constrained by limited power budgets.

The paper categorizes power management techniques into four principal groups:

1. Dynamic Voltage and Frequency Scaling (DVFS) and Power-Aware Scheduling: DVFS, a widely recognized method, adjusts the voltage and frequency based on power and performance needs, although it has diminishing returns due to increased leakage energy and the prevalence of multi-core processors. The paper reviews numerous algorithms that optimize DVFS alongside task scheduling to maximize energy savings without compromising deadlines. Techniques utilize offline profiling and runtime adaptation to navigate the performance-energy trade-off efficiently.
2. Power Mode Management (PMM): PMM leverages the availability of various operating modes to save energy during idle periods. The paper discusses strategies for transitioning systems between normal and low-power modes effectively, ensuring that energy savings are achieved while maintaining performance criteria. These methods include computing break-even times and utilizing historical task statistics for adaptive mode transitioning.
3. Microarchitectural Techniques: These strategies aim to save energy in specific system components, such as main memory, cache, or entire memory hierarchies. The reviewed techniques dynamically reconfigure components, exploiting workload variation, or leveraging application-specific characteristics. Notable approaches include cache reconfiguration and the employment of alternative memory architectures, like scratchpad memory.
4. Unconventional Cores: The paper explores the use of DSPs, FPGAs, and GPUs as alternatives to traditional CPU cores, emphasizing their role in enhancing energy efficiency. Compared to CPUs, these cores provide significant performance benefits that translate into better energy-performance ratios for specific applications, such as signal processing or high-definition video processing.

The survey provides a robust classification of state-of-the-art approaches and reflects on the implications of energy efficiency in embedded systems design. It suggests that future mobile computing platforms will demand significant improvements in energy management due to high-performance requirements in thought-intensive applications. The integrated methodologies detailed here offer a foundation for new research avenues that will enhance the sophistication and efficiency of embedded systems, offering practical benefits in terms of longevity, reliability, and adoption across diverse domains.

The paper, by presenting a comprehensive overview of existing research in power management, signifies the importance of continued advancements in this area. Future research could explore deeper integration of unconventional computing platforms, refine microarchitectural solutions, and improve system-level power management, aiming for further efficiency gains in energy-constrained environments.