微型電腦叢集化建置節能伺服器資料中心

壽大衛 ； 林岑威

《TANET2016 臺灣網際網路研討會》 2016 (2016/11) Pp. 612-615-110

https://doi.org/10.6679/TANET.2016.110

摘要

本研究以多部微型電腦叢集化搭配開源作業系統製作伺服器基礎建置群集。同時透過容器(Linux Container)技術，將作業系統上層應用程式虛擬化，由開源系統軟體做應用叢系統管理。此建構開發出可以快速部屬、建置、可靠性及可用性之伺服器基礎建置，以提供低能耗、低成本之綠能化伺服器群集。由於本系統架構因降低成本與節能考量，使用的是ARM CPU 的微型電腦，透過網路交換器進行並聯組成實體伺服器群集，而群集也統一採用開源作業系統組成作為低成本的運算資料中心。本系統因統一由Docker Swarm 管理容器達到叢集化服務，也因開源軟體發展突飛猛進使得軟體定義資料中心(SDDC)不在是x86 cpu 硬體和VMware 底層虛擬化的唯一技術，反觀是開源軟體讓更多企業和組織能發展最適合自己的資料中心。本研究能證明微型電腦可以快速部屬成資料中心已提供低能源消耗的運算之中心外，並同時偵測應用層虛擬化容器技術的容器數量與電壓電流的關係對應，產生的相對演算方法，以此方法可以應用在發展快速的物聯網資料中心。

關鍵字

LXC ； Docker Swarm ； RaspberryPi ； Cluster

並列摘要

In this paper, more than a cluster of miniature computerswith open source operating systems to create server based build cluster. At the same time through the Container technology, the upper layer application operating system virtualization, open source software made by the application cluster system management. This construct can quickly develop subordinates, build, server reliability and availability of the foundation, in order to provide low-power, low-cost green energy of the server cluster. As the result of the system architecture to reduce cost and energy-saving considerations, using ARM micro-computer, server clusters in parallel constituent entities through the network switch, and the cluster has a unified open source operating system components as a low-cost computing data centers. The system is due to reach a unified cluster management service by the Docker Swarm container, but also because of the rapid development of open-source software so that software-defined data center (SDDC) is not x86 cpu underlying hardware and VMware virtualization technology only, the other hand is open-source software to allow more businesses and organizations to develop the most suitable for their own data centers. This study proves that the microcomputer can quickly subordinate to data center computing center has provided low energy consumption, the relationship and simultaneously detect the corresponding application layer virtualization technology container container number and the voltage and current, resulting in a relative calculation method in order this method can be used in fast-growing data center networking.

並列關鍵字

LXC ； Docker Swarm ； RaspberryPi ； Cluster