Smart Societies, Infrastructure, Technologies and Applications. First International Conference, SCITA 2017, Jeddah, Saudi Arabia, November 27–29, 2017, Proceedings

Research Article

Power Efficiency of a SBC Based Hadoop Cluster

Download
196 downloads
  • @INPROCEEDINGS{10.1007/978-3-319-94180-6_7,
        author={Basit Qureshi and Anis Koubaa},
        title={Power Efficiency of a SBC Based Hadoop Cluster},
        proceedings={Smart Societies, Infrastructure, Technologies and Applications. First International Conference, SCITA 2017, Jeddah, Saudi Arabia, November 27--29, 2017, Proceedings},
        proceedings_a={SCITA},
        year={2018},
        month={7},
        keywords={Hybrid cloud computing Low cost cluster Power efficiency},
        doi={10.1007/978-3-319-94180-6_7}
    }
    
  • Basit Qureshi
    Anis Koubaa
    Year: 2018
    Power Efficiency of a SBC Based Hadoop Cluster
    SCITA
    Springer
    DOI: 10.1007/978-3-319-94180-6_7
Basit Qureshi1,*, Anis Koubaa,*
  • 1: Prince Sultan University
*Contact email: qureshi@psu.edu.sa, akoubaa@psu.edu.sa

Abstract

Security, reliability and energy efficiency are key requirements in any smart cities applications. In this paper, we investigate the use of a power efficient cloud cluster composed of Single Board Computers (SBC). A Low-Cost Cloud Computing Cluster (LoC4) is designed using Raspberry Pi and Odroid Xu4 Single Board Computers. The LoC4 cluster is thoroughly tested for performance and power efficiency using TeraGen and TeraSort benchmarks that are used primarily for stress testing the Hadoop cluster. Results show that Odroid Xu4 computers consume more energy compared to RPi computers, however the overall energy consumption in the cluster for completion of a task was 91% and 72% better for Odroid Xu4 with workloads of 800 MB and 400 MB respectively.