GPU集群能耗优化控制模型研究

doi:10.3969/j.issn.0372-2112.2015.10.004

PDF(1481 KB)

电子学报 ›› 2015, Vol. 43 ›› Issue (10) : 1904-1910. DOI: 10.3969/j.issn.0372-2112.2015.10.004

学术论文

GPU集群能耗优化控制模型研究

王海峰^1,2, 曹云鹏^1,2

作者信息 +

Power Consumption Optimization Control Model of GPU Clusters

WANG Hai-feng^1,2, CAO Yun-peng^1,2

Author information +

文章历史 +

摘要

随着大数据技术的发展,GPU集群作为一种高效的并行系统被应用到大规模数据实时计算中.能量是实时计算时重要的资源,GPU集群的能耗优化及实时消减成为一个具有挑战性的问题.从集群全局角度引入模型预测控制策略,并建立闭环反馈机制的多输入多输出控制器.通过调整计算频率和改变活跃流多处理器来改变能耗状态,利用反馈和滚动优化机制完成对未来的控制预判,实现消减冗余能耗的目标.实验表明:控制模型的精度和节能效果优于基准模型,而且具有较好的稳定性,适合应用到大规模数据实时计算中.

Abstract

With the development of Big Data technology GPU cluster as a high efficiency parallel system applies into the Large-scale data computing field.Energy is a significant computation resource.So power consumption optimization control and capping in real-time becomes a challenge issue.The Model Prediction Control strategy is introduced and a Multi-Input Multi-Output controller is built by using a closed loop feedback principle from the whole cluster perspective.Power consumption status is changed by scaling frequency and adjusting active stream multi-processors.Then the feedback and the periodic optimization mechanisms can predict the control behaviors in the future control cycles.This achieves the goal that reduces redundancy energy.The results demonstrate that the proposed model has more accuracy and comsumes less energy than the others.And it has better control stability.So it has better adaptability and obvious advantage in the Large-scale data real-time computing.

导出引用

王海峰, 曹云鹏. GPU集群能耗优化控制模型研究[J]. 电子学报, 2015, 43(10): 1904-1910. https://doi.org/10.3969/j.issn.0372-2112.2015.10.004

WANG Hai-feng, CAO Yun-peng. Power Consumption Optimization Control Model of GPU Clusters[J]. Acta Electronica Sinica, 2015, 43(10): 1904-1910. https://doi.org/10.3969/j.issn.0372-2112.2015.10.004

中图分类号： TP302.7

参考文献

[1] 杨正龙,金林,李蔚清.基于GPU的图形电磁计算加速算法[J].电子学报,2007,35(6):1056-1060. Yang Zheng-long,Jin Lin,Li Wei-qing.Accelerated GRECObased on GPU[J].Acta Electronica Sinica,2007,35(6):1056-1060.(in Chinese)
[2] 刘勇鹏,王锋,卢凯,等.面向异构并行计算系统的流水线式压缩检查点[J].电子学报,2012,40(2):223-229. Liu Yong-peng,Wang Feng,Lu Kai,et al.Pipelinedcompressed checkpointing for heterogeneous systems[J].Acta Electronica Sinica,2012,40(2):223-229.(in Chinese)
[3] Horvath T,Abdelzaher T,Shadron K,et al.Dynamic voltage scaling in multitier web servers with end-to-end delay control[J].IEEE Transaction on Computers,2007,56(4):444-458.
[4] Bertini L,Julius C B,Mosse D.Power optimization for dynamic configuration in heterogeneous web server clusters[J].Journal of Sysems and Software,2010,83(4):585-598.
[5] Wang Xiao-rui.Coordinating power control and performance management for virtualized server clusters[J].IEEE Transaction on Parallel and Distributed System,2011,22(2):245-259.
[6] 林一松,杨学军,唐滔,等.一种基于并行度分析模型的GPU功耗优化技术[J].计算机学报,2011,34(4):705-716. Lin Yi-song,Yang Xue-jun,Tang Tao,et al.A GPU low-power optimization based on parallelism analysis model[J].Chinese Journal of Computers,2011,34(4):705-716.(in Chinese)
[7] 林一松,杨学军,唐滔,等.一种基于关键路径分析的CPU-GPU异构系统综合能耗优化方法[J].计算机学报,2012,35(1):123-133. Lin Yi-song,Yang Xue-jun,Tang Tao,et al.Anintegrated energy optimization approach for CPU-GPU heterogeneous system based on critical path analysis[J].Chinese Journal of Computers,2012,35(1):123-133.(in Chinese)
[8] 赵霞,郭耀,雷志勇,等.基于模拟器的嵌入式操作系统能耗估算与分析[J].电子学报,2008,36(2):209-215. Zhao Xia,Guo Yao,Lei Zhi-yong,et al.Estimation andanalysis of embedded operating system energy consumption[J].Acta Electronica Sinica,2008,36(2):209-215.(in Chinese)
[9] 王海峰,陈庆奎.静态程序切片的GPU通用计算功耗预测模型[J].软件学报,2013,24(8):1746-1760. Wang Hai-feng,Chen Qing-kui.Powerconsumption prediction model of general-purpose computing GPU with static program slicing[J].Journal of Software,2013,24(8):1746-1760.(in Chinese)
[10] Wang Hai-feng,Chen Qing-kui.Power estimating model and analysis of general programming on GPU[J].Journal of Software,2012,7(5):1164-1170.
[11] Tang Qing-hui,Gupta S K,Georgios V.Energy-efficient thermal-aware task scheduling for homogeneous high-performance computing data centers:A cyber-physical approach[J].IEEE Transaction on Parallel and Distributed System,2008,19(11):1458-1472.
[12] 徐祖华,赵均,钱积新.基于多自由度性能指标的模型预测控制算法[J].电子学报,2008,36(5):906-909. Xu Zu-hua,Zhao Jun,Qian Ji-xin.Animproved model predictive control algorithm based on multi-degree-of freedom performance index[J].Acta Electronica Sinica,2008,36(5):906-909.(in Chinese)
[13] Yang Xue-jun,Yan Xiao-bo,Xing Zuo-cheng,et al.Fei teng 64 stream processing system:architecture,compiler,and programming[J].IEEE Transtractions on Parallel and Distributed Systems,2009,20(8):1142-1156.
[14] Hong S,Kim H.An analytical model for a gpu architecture with memory-level and thread-level parallelism awareness[A].Proceedings of ISCA'09[C].New York:ACM,2009.152-163.
[15] Hong S,Kim H.An integrated GPU power and performance model[A].Proceedings of ISCA'10[C].New York:ACM,2010.280-289.
[16] 王海峰,陈庆奎.图形处理器通用计算关键技术研究综述[J].计算机学报,2013,36(4):757-772. Wang Hai-feng,Chen Qing-kui.Generalpurpose computing of graphics processing unit:a survey[J].Chinese Journal of Computers,2013,36(4):757-772.(in Chinese)