標題: 在混合負載即時系統下之高能源效率動態電壓調整演算法
Energy Efficient Dynamic Voltage Scaling for Mixed Workload Real-time Systems
作者: 陳蒸民
Jheng-Ming Chen
王國禎
Kuochen Wang
網路工程研究所
關鍵字: 實際工作量;任務間動態電壓調整方法;混合負載即時系統;寬裕時間;最差執行時間;Actual workload;inter-task dynamic voltage scaling;mixed workload real-time system;slack time;worst case execution time
公開日期: 2006
摘要: 近年來,由於無線通訊的快速成長,使個人數位助理與手機等手持裝置的能源消耗成為一個值得重視的研究問題。儘管有許多動態電壓調整演算法被提出來降低即時系統上的能源消耗,但絕大部分的動態電壓調整演算法僅針對只有週期性工作的即時系統,甚少動態電壓調整演算法適用於週期性工作與非週期性工作同時存在的混合負載即時系統。一個混合負載即時系統下的動態電壓調整演算法,不僅要節省能源消耗,同時也必須考慮到非週期性工作的反應時間。本論文提出一個在速率單調(Rate Monotonic)排程法下針對混合負載即時系統之基於工作需求估計寬裕時間(slack time)的高能源效率動態電壓調整演算法(WSS)。WSS估計寬裕時間是依據非週期性工作排程伺服器的執行行為和原本只適用於純週期性工作即時系統下的短期工作需求分析法。除此之外,當非週期性的工作量大於非週期性工作排程伺服器所能服務的量時,WSS會利用所估計的寬裕時間來提前服務非週期性工作,因此可縮短非週期性工作的反應時間。模擬結果顯示,我們所提出的方法相較於現存的SNRT (BSS) 方法,節省了23% (13%) 的能源消耗,縮短了38% (31%) 的非週期性工作的反應時間,同時也減少 52% (40%) 的能源消耗與反應時間乘積。
Recently the wireless communication is rapidly growing up and the energy consumption on mobile devices, such as personal digital assistants (PDAs) and cellular phones, becomes a critical issue. In spite of numerous inter-task dynamic voltage scaling (DVS) algorithms have been brought up for energy saving of real-time systems with only periodic tasks or only aperiodic tasks, few of them were aimed at the mixed workload of periodic tasks and aperiodic tasks. A DVS algorithm for mixed workload real-time systems should not only focus on energy saving, but also consider low response time of aperiodic tasks. In this thesis, we develop an on-line energy efficient scheduling, Work-demand-based Slack-Stealing scheme (WSS), to reduce CPU energy consumption for mixed workload real-time systems under the Rate Monotonic (RM) scheduling policy. The WSS calculates the available slack time by the execution behaviors of scheduling servers for aperiodic tasks and by short-term work-demand analysis, which was originally designed for real-time systems with periodic tasks only. Moreover, the WSS also utilizes the concept of slack stealing to service aperiodic tasks and to reduce the response time when the actual workload of aperiodic tasks is close to the server utilization. Simulation results shows that the proposed WSS can effectively reduce the energy consumption, response time, and energy consumption * response time, in average, by 23%, 38%, 52% compared to the SNRT and 13%, 31% , 40% compared to the BSS, respectively.
URI: http://140.113.39.130/cdrfb3/record/nctu/#GT009456511
http://hdl.handle.net/11536/82176
Appears in Collections:Thesis


Files in This Item:

  1. 651101.pdf

If it is a zip file, please download the file and unzip it, then open index.html in a browser to view the full text content.