Yung-Hsiang Lu

A Survey of Computation Offloading for Mobile Systems

Mobile systems have limited resources, such as battery life, network bandwidth, storage capacity, and processor performance. These restrictions may be alleviated by computation offloading: sending heavy computation to resourceful servers and receiving the results from these servers. Many issues related to offloading have been investigated in the past decade. This survey paper provides an overview of the background, techniques, systems, and research areas for offloading computation. We also describe directions for future research.

Comparing system level power management policies

Reducing power consumption is a challenge to system designers. Portable systems, such as laptop computers and personal digital assistants (PDAs), draw power from batteries, so reducing power consumption extends their operating times. For desktop computers or servers, high power consumption raises temperature and deteriorates performance and reliability. Soaring energy prices and rising concern about the environmental impact of electronics systems further highlight the importance of low power consumption. Power reduction techniques can be classified as static and dynamic. Static techniques, such as synthesis and compilation for low power, are applied at design time. In contrast, dynamic techniques use runtime behavior to reduce power when systems are serving light workloads or are idle. These techniques are known as dynamic power management (DPM). DPM can be achieved in different ways; for example, dynamic voltage scaling (DVS) changes supply voltage at runtime as a method of power management. Here, we use DPM specifically for shutting down unused I/O devices. We built an experimental environment on a laptop computer running Microsoft Windows. We implemented existing power management policies and quantitatively compared their effects on power saving and performance degradation.

Dynamic power management for nonstationary service requests

Dynamic power management (DPM) is a design methodology aimed at reducing power consumption of electronic systems by performing selective shutdown of idle system resources. The effectiveness of a power management scheme depends critically on accurate modeling of service requests and on computation of the control policy. In this work, we present an online adaptive DPM scheme for systems that can be modeled as finite-state Markov chains. Online adaptation is required to deal with initially unknown or nonstationary workloads, which are very common in real-life systems. Our approach moves from exact policy optimization techniques in a known and stationary stochastic environment and extends optimum stationary control policies to handle the unknown and nonstationary stochastic environment for practical applications. We introduce two workload learning techniques based on sliding windows and study their properties. Furthermore, a two-dimensional interpolation technique is introduced to obtain adaptive policies from a precomputed look-up table of optimum stationary policies. The effectiveness of our approach is demonstrated by a complete DPM implementation on a laptop computer with a power-manageable hard disk that compares very favorably with existing DPM schemes.

Quantitative comparison of power management algorithms

Dynamic power management saves power by shutting down idle devices. Several management algorithms have been proposed and demonstrated to be effective in certain applications. We quantitatively compare the power saving and performance impact of these algorithms on hard disks of a desktop and notebook computers. This paper has three contributions. First, we build a framework in Windows NT to implement power managers running realistic workloads and directly interacting with users. Second, we define performance degradation that reflects user perception. Finally, we compare power saving and performance of existing algorithms and analyze the difference.

A case study of mobile robot's energy consumption and conservation techniques

Mobile robots are used in many applications, such as carpet cleaning, pickup and delivery, search and rescue, and entertainment. Energy limitation is one of the most important challenges for mobile robots. Most existing studies on mobile robots focus on motion planning to reduce motion power. However, motion is not the only power consumer. In this paper, we present a case study of a mobile robot called Pioneer 3DX. We analyze the energy consumers. We build power models for motion, sonar sensing and control based on experimental results. The results show that motion consume less than 50% power on average. Therefore, it is important to consider the other components in energy-efficient designs. We introduce two energy-conservation techniques: dynamic power management and real-time scheduling. We provide several examples showing how these techniques can be applied to robots. These techniques together with motion planning provide greater opportunities to achieve better energy efficiency for mobile robots. Although our study is based on a specific robot, the approach can be applied to other types of robots

Adaptive computation offloading for energy conservation on battery-powered systems

This paper considers the problem of extending the battery lifetime for a portable computer by off loading its computation to a server. Depending on the inputs, computation time for different instances of a program can vary significantly and they are often difficult to predict. Different from previous studies on computation off loading, our approach does not require estimating the computation time before the execution. We execute the program initially on the portable client with a timeout. If the computation is not completed after the timeout, it is off loaded to the server. We first set the timeout to be the minimum computation time that can benefit from off loading. This method is proved to be 2- competitive. We further consider collecting online statistics of the computation time and find the statistically optimal timeout. Finally, we provide guidelines to construct programs with computation off loading. Experiments show that our methods can save up to 17% more energy than existing approaches.

Adaptive hard disk power management on personal computers

Dynamic power management can be effective for designing low-power systems. In many systems, requests are clustered into sessions. This paper proposes an adaptive algorithm that can predict session lengths and shut down components between sessions to save power. Compared to other approaches, simulations show that this algorithm can reduce power consumption in hard disks with less impact on performance or reliability.

Low-power task scheduling for multiple devices

Power management saves power by shutting down idle devices. These devices often serve requests from concurrently running tasks. Ordering task execution can adjust the lengths of idle periods and exploit better opportunities for power management. This paper presents an on-line low-power scheduling algorithm for multiple devices. Simulations show that it can save up to 33% power and reduce 40% state-transition delays. This algorithm is robust under imperfect knowledge of future requests and timing constraints; therefore, it is applicable to interactive systems.

Sensor replacement using mobile robots

Sensor replacement is important for sensor networks to provide continuous sensing services. Upon sensor node failures, holes (uncovered areas) may appear in the sensing coverage. Existing approaches relocate redundant nodes to fill the holes and require all or most sensor nodes to have mobility. However, mobility equipment is expensive while technology trends are scaling sensors to be smaller and cheaper. In this paper, we propose to use a small number of mobile robots to replace failed sensors for a large-scale static sensor network. We study algorithms for detecting and reporting sensor failures and coordinating the movement of robots that minimize the motion energy of mobile robots and the messaging overhead incurred to the sensor network. A manager receives failure reports and determines which robot to handle a failure. We study three algorithms: a centralized manager algorithm, a fixed distributed manager algorithm, and a dynamic distributed manager algorithm. Our analysis and simulations show that: (a) the centralized and the dynamic distributed algorithms have lower motion overhead than the fixed distributed algorithm; (b) the centralized algorithm is less scalable than the two distributed manager algorithms, and (c) the two distributed algorithms have higher messaging cost than the centralized algorithm. Hence, the optimal choice of the coordination algorithm depends on the specific scenarios and objectives being optimized.

Software controlled power management

Reducing power consumption is critical in many system designs. Dynamic power management is an effective approach to decrease power without significantly degrading performance. Power management decisions can be implemented in either hardware or software. A recent trend on personal computers is to use software to change hardware power states. This paper presents a software architecture that allows system designers to investigate power management algorithms in a systematic fashion through a template. The architecture exploits the Advanced Configuration and Power Interface (ACPI), a standard for hardware and software. We implement two algorithms for controlling the power states of a hard disk on a personal computer running Microsoft Windows. By measuring the current feeding the hard disk, we show that the algorithms can save up to 25% more energy than the Windows power manager. Our work has two major contributions: a template for software-controlled power management and experimental comparisons of management algorithms for a hard disk.