Electrical & Computer Engineering Research Works
Permanent URI for this collectionhttp://hdl.handle.net/1903/1658
Browse
9 results
Search Results
Item Dual-Processor Design of Energy Efficient Fault-Tolerant System(IEEE, 2006-09) Hua, Shaoxiong; Pari, Pushkin R.; Qu, GangA popular approach to guarantee fault tolerance in safety-critical applications is to run the application on two processors. A checkpoint is inserted at the comple- tion of the primary copy. If there is no fault, the sec- ondary processor terminates its execution. Otherwise, should the fault occur, the second processor continues and completes the application before its deadline. In this paper, we study the energy efficiency of such dual- processor system. Specifically, we first derive an opti- mal static voltage scaling policy for single periodic task. We then extend it to multiple periodic tasks based on worst case execution time (WCET) analysis. Finally, we discuss how to further reduce system’s energy con- sumption at run time by taking advantage of the actual execution time which is less than the WCET. Simula- tion on real-life benchmark applications shows that our technique can save up to 80% energy while still provid- ing fault tolerance.Item Power Minimization Techniques on Distributed Real-Time Systems by Global and Local Slack Management(IEEE, 2005-01) Hua, Shaoxiong; Qu, GangItem QOS-DRIVEN SCHEDULING FOR MULTIMEDIA APPLICATIONS(IEEE, 2004-05) Hua, Shaoxiong; Qu, Gang;Multimedia applications have intrinsic quality of service (QoS) requirements that may not be captured by the simple traditional completion ratio model. We have proposed a new quantitative QoS metric based on task completion ratio while differentiating firm and soft deadlines and taking data dependency into consideration. Using the decoding of MPEG movies as an example, we have shown that the proposed QoS metric is much better than completion ratio in measuring the quality of presentation (QoP) of the movies. Based on the new QoS metric, we present a set of new online algorithms that outperform popular scheduling algorithms (such as EDF, FCFS, and LETF) and enhance QoP significantly, particularly when the system is overloaded. All the proposed online algorithms have low computation overhead and can be easily integrated into real-time operating systems to improve multimedia embedded system’s performance and/or to save system resources.Item Approaching the Maximum Energy Saving on Embedded Systems with Multiple Voltages(IEEE, 2003-11) Hua, Shaoxiong; Qu, Gang;Dynamic voltage scaling (DVS) is arguably the most effective energy reduction technique. The multiple-voltage DVS systems, which can operate only at pre-determined discrete voltages, are practical and have been well studied. However, one important unsolved problem is how many levels and at which values should voltages be implemented on a multiple-voltage DVS system to achieve the maximum energy saving. We refer this as the voltage set-up problem. In this paper, (1) we derive analytical solutions for dual-voltage system. (2) For the general case that does not have analytic solutions, we develop efficient numerical methods. (3) We demonstrate how to apply the proposed algorithms on system design. (4) Interestingly, the experimental results suggest that the multiple-voltage DVS system, when the voltages are set up properly, can reach DVS technique’s full potential in energy saving. Specifically, on the design of an ad hoc application specific system and the design of the MPEG video encoder, we find that the best single-voltage systems consume 150% and 20% more energy than the tight theoretical lower bounds, respectively. However, our approach gives dual-, 3-, and 4-voltage DVS system settings that are only 17.6%, 4.9%, and 2.6% for the ad hoc system, and 4.0%, 1.1%, and 0.2% for the MPEG video encoder, over the same lower bounds.Item Exploring the Probabilistic Design Space of Multimedia Systems(IEEE, 2003-06) Hua, Shaoxiong; Qu, Gang; Bhattacharyya, Shuvra S.In this paper, we propose the novel concept of probabilistic design for multimedia systems and a methodology to quickly explore such design space at an early design stage. The probabilistic design is motivated by the challenge of how to design, but not over-design, multimedia embedded systems while systematically incorporating such application’s performance requirements, uncertainties in execution time, and tolerance for reasonable execution failures. Our goal is to bridge the gap between real-time analysis and embedded software implementation for rapid and economic (multimedia) system prototyping. Our method takes advantage of multimedia system’s unique features mentioned above to relax the rigid hardware requirements for software implementation and eventually avoid over-designing the system.Item Energy Reduction Techniques for Multimedia Applications with Tolerance to Deadline Misses(IEEE, 2003-06) Hua, Shaoxiong; Qu, Gang; Bhattacharyya, Shuvra S.Many embedded systems such as PDAs require processing of the given applications with rigid power budget. However, they are able to tolerate occasional failures due to the imperfect human visual/auditory systems. The problem we address in this paper is how to utilize such tolerance to reduce multimedia system’s energy consumption for providing guaranteed quality of service at the user level in terms of completion ratio. We explore a range of offline and on-line strategies that take this tolerance into account in conjunction with the modest non-determinism in application’s execution time. First, we give a simple best-effort approach that achieves the maximum completion ratio; then we propose an enhanced on-line best-effort energy minimization (BEEM) approach and a hybrid offline/on-line minimumeffort (O2ME) approach. We prove that BEEM maintains the maximum completion ratio while consuming the provably least amount of energy and O2ME guarantees the required completion ratio statistically. We apply both approaches to a variety of benchmark task graphs, most from popular DSP applications. Simulation results show that significant energy savings (38% for BEEM and 54% for O2ME, both over the simple best-effort approach) can be achieved while meeting the required completion ratio requirements.Item A New Quality of Service Metric for Hard/Soft Real-Time Applications(IEEE, 2003-04) Hua, Shaoxiong; Qu, GangReal-time applications often have mixed hard and soft deadlines, can be preempted subject to the cost of context switching or the restart of computation, and have various data dependency. The simple but widely used task completion ratio, as the Quality of Service (QoS) metric, does not capture these characteristics and can not reflect user perceived QoS well. In this paper, we propose a new quantitative QoS metric, which is based on task completion ratio but differentiates hard and soft deadlines and models data dependency as well. Basically, it assigns different weights to hard and soft deadline tasks, penalizes late soft task completion, and measures the tasks affected by any dropped tasks. We apply popular online schedulers, such as EDF (earliest deadline first), FCFS (first come first serve), and LETF (least execution time first), on a set of simulated MPEG movies at the frame level and for each application compare the new QoS measurement, traditional completion ratio with the “real” completion ratio which considers the number of correctly decoded frames and has been mapped to the user perceived QoS well. Experimental results show that our proposed QoS metric can reflect real life QoS much better than the traditional one.Item Voltage Set-up Problem for Embedded Systems with Multiple Voltages.(IEEE, 2005-07) Hua, Shaoxiong; Qu, GangItem QoP-Driven Scheduling for MPEG Video Decoding(IEEE, 2003-11) Hua, Shaoxiong; Qu, GangMPEG video decoding algorithm has been embedded into many consumer electronic products. In this paper, we demonstrate that the completion ratio (CR) does not represent well the quality of presentation (QoP) of MPEG video. We then propose a new QoP metric, which 1) is based on frame completion ratio but 2) differentiates firm and soft deadlines and also 3) considers the frame dependency. We show that, on a set of simulated MPEG movies, the proposed QoP metric measures the QoP of the movies much better than the completion ratio. We then present a set of online scheduling algorithms that enhance QoP significantly, particularly for overloaded systems.