energy.pdf - U.P.B Sci Bull Series C Vol 79 Iss 1 2017 ISSN...

Info icon This preview shows pages 1–3. Sign up to view the full content.

View Full Document Right Arrow Icon
U.P.B. Sci. Bull., Series C, Vol. 79, Iss. 1, 2017 ISSN 2286-3540 EVALUATION OF A LOW-POWER HADOOP CLUSTER BASED ON THE ZYNQ ARM-FPGA SOC Ovidiu PLUGARIU 1 , Alexandru CALIN 2 , and Lucian PETRICA 3 Distributed computing is important to many web and scientific ap- plications. The quest for lower power dissipation and energy consumption in distributed applications has led to the implementation of ARM proces- sor based distributed systems. The advent of integrated accelerators such as Field Programmable Gate Arrays (FPGA) promises to provide further reductions in power, but it is yet unclear to the scientific community how FPGAs should be utilized to maximize their benefit to the distributed sys- tem. To answer this research question, in this paper we evaluate a ARM- FPGA distributed system which utilizes the Xilinx Zynq SoC as processor and Apache Hadoop as a programming environment and task scheduler. We evaluate the system on industry-standard benchmarks: MRBench, TestDF- SIO, Wordcount. Compared to a similar system with x86 processors, the ARM-based system dissipates 5 times less power. Additionally, we observe that all applications are IO bound on the ARM system, which indicates that the most efficient use for FPGAs in such a system is for offloading IO compression algorithms. Keywords: Hadoop, ARM, distributed computing 1. Introduction All modern web-based applications, ranging from web analytics, to search, and social network infrastructure, must process thousands of requests per sec- ond. These applications therefore require vast amounts of computing power, i.e. hundreds to thousands of co-located, networked servers which work si- multaneously and collaboratively to process information and generate results. This type of computing infrastructure is called a cluster, and the process of collaborative task solving is called distributed computing. Modern distributed computing occurs in data centers, facilities which house interconnected servers and associated components, such as networking and storage systems [1]. As such, data-center efficiency dictates the profitability of most on-line businesses. The majority of data-center operational expenditure is the cost for electrical 1 PhD Student, DCAE Department, University “Politehnica” of Bucharest, Romania 2 PhD Student, DCAE Department, University “Politehnica” of Bucharest, Romania 3 Lector, DCAE Department, University “Politehnica” of Bucharest, Romania, e-mail: [email protected] 125
Image of page 1

Info icon This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document Right Arrow Icon
126 O. Plugariu, A. Calin, and L. Petrica energy for servers and cooling equipment, which may be reduced by improving the energy-efficiency of the server processors utilized. Low-power computing is an area of great interest for industry and academia because it enables reductions in the cost of distributed computing in data- centers. Low-power servers based on ARM processors have the potential of increased efficiency - more operations per Watt and per Joule, compared to servers which utilize processors of the x86 and PowerPC architectures, which are faster at the cost of significantly greater power dissipation [2]. With the ad-
Image of page 2
Image of page 3
This is the end of the preview. Sign up to access the rest of the document.

{[ snackBarMessage ]}

What students are saying

  • Left Quote Icon

    As a current student on this bumpy collegiate pathway, I stumbled upon Course Hero, where I can find study resources for nearly all my courses, get online help from tutors 24/7, and even share my old projects, papers, and lecture notes with other students.

    Student Picture

    Kiran Temple University Fox School of Business ‘17, Course Hero Intern

  • Left Quote Icon

    I cannot even describe how much Course Hero helped me this summer. It’s truly become something I can always rely on and help me. In the end, I was not only able to survive summer classes, but I was able to thrive thanks to Course Hero.

    Student Picture

    Dana University of Pennsylvania ‘17, Course Hero Intern

  • Left Quote Icon

    The ability to access any university’s resources through Course Hero proved invaluable in my case. I was behind on Tulane coursework and actually used UCLA’s materials to help me move forward and get everything together on time.

    Student Picture

    Jill Tulane University ‘16, Course Hero Intern