Conference Paper

Green Supercomputing in a Desktop Box

Dept. of Comput. Sci., Virginia Tech., Blacksburg, VA
DOI: 10.1109/IPDPS.2007.370542 Conference: Parallel and Distributed Processing Symposium, 2007. IPDPS 2007. IEEE International
Source: IEEE Xplore

ABSTRACT The advent of the Beowulf cluster in 1994 provided dedicated compute cycles, i.e., supercomputing for the masses, as a cost-effective alternative to large supercomputers, i.e., supercomputing for the few. However as the cluster movement matured, these clusters became like their large-scale supercomputing brethren - a shared (and power-hungry) datacenter resource that must reside in a actively-cooled machine room in order to operate properly. The above observation, coupled with the increasing performance gap between the PC and supercomputer, provides the motivation for a "green supercomputer" in a desktop box. Thus, this paper presents and evaluates such an architectural solution: a 12-node personal desktop supercomputer that offers an interactive environment for developing parallel codes and achieves 14 Gflops on Linpack but sips only 185 watts of power at load - all this in the approximate form factor of a Sun SPARCstation 1 pizza box.

  • [Show abstract] [Hide abstract]
    ABSTRACT: This paper analyses the power consumption of hybrid computation on embedded architectures with an available GPU. Novel efficiency metrics are obtained using a well-known benchmark process based on the Fourier transform as computing work load. The measurement process is arranged in order to obtain specific power data for each hardware configuration, varying the data size and number of computation threads, disabling the GPU, mixing the power computation of CPU/GPU, etc. The resulting data may be of interest for new applications and cluster development (i.e. Beowulf clusters) based on low power devices, such as the Beobot project.
    Proceedings of the 2012 IEEE 14th International Conference on High Performance Computing and Communication & 2012 IEEE 9th International Conference on Embedded Software and Systems; 01/2012
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Reducing energy consumption for high end computing can bring various benefits such as reducing operating costs, increasing system reliability, and environmental respect. This paper aims to develop scheduling heuristics and to present application experience for reducing power consumption of parallel tasks in a cluster with the Dynamic Voltage Frequency Scaling (DVFS) technique. In this paper, formal models are presented for precedence-constrained parallel tasks, DVFS-enabled clusters, and energy consumption. This paper studies the slack time for non-critical jobs, extends their execution time and reduces the energy consumption without increasing the task’s execution time as a whole. Additionally, Green Service Level Agreement is also considered in this paper. By increasing task execution time within an affordable limit, this paper develops scheduling heuristics to reduce energy consumption of a tasks execution and discusses the relationship between energy consumption and task execution time. Models and scheduling heuristics are examined with a simulation study. Test results justify the design and implementation of proposed energy aware scheduling heuristics in the paper.
    Future Generation Computer Systems 09/2013; 29(7):1661-1670. · 2.64 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Energy management has become increasingly necessary in large-scale cloud data centers to address high operational costs and carbon footprints to the environment. In this work, we combine three management techniques that can be used to control cloud data centers in an energy-efficient manner: changing the number of virtual machines, the number of cores, and scaling the CPU frequencies. We present a feedback controller that determines an optimal configuration to minimize energy consumption while meeting performance objectives. The controller can be configured to accomplish these goals in a stable manner, without causing large oscillations in the resource allocations. To meet the needs of individual applications under different workload conditions, the controller parameters are automatically adjusted at runtime based on a system model that is learned online. The potential of the proposed approach is evaluated in a video encoding scenario. The results show that our combined approach achieves up to 34% energy savings compared to the constituent approaches—core change, virtual machine change, and CPU frequency change policies, while meeting the performance target.
    Sustainable Computing: Informatics and Systems. 08/2014;


Available from