NCTU CStack：OpenStack 與 Ceph 的整合與應用

Abstract

雲端運算是近年來非常熱門的話題，隨著需要處理的資料愈來愈龐大，各個機構所設置的資訊中心也愈來愈複雜。因此，關於資料如何安全的保存，以及硬體設備如何有效的利用，也變成非常重要的議題。 在資料安全性方面，Ceph為一新崛起的分散式儲存系統，設計之初即把把底層的硬體離線維護視為常態，並不會因為小數量的機器故障而造成整個系統停擺，進而提供了很好的資料安全性。 在硬體利用率方面，虛擬化是一個很好的解決方式。傳統的虛擬化軟體可以在一台機器上開啟多台虛擬設備，但若要管理一堆虛擬機的集合，分配不同的虛擬機給不同的單位使用，以及處理虛擬機與實體機的網路連線等，則非常麻煩。因此，虛擬化的主要廠商皆推出了相對應的產品，如VMware ESXi搭配vCenter來建構虛擬化的資料中心，或是Microsoft Windows Server的Hyper-V以及Citrix XenServer等，都可以快速的提供企業自己的私有雲服務，但是昂貴的授權費用，以及對硬體設備的規格需求相對嚴苛的限制，也讓不少機構望之卻步。 OpenStack為近年由NASA和Rackspace共同開發的開放原始碼軟體，可提供任何人建立免費的雲端運算服務。我們利用了OpenStack可快速建立私有雲的特性，搭配了Ceph分散式儲存系統，在交通大學資訊技術服務中心建立了一個名為NCTU CStack的系統，提供免費且高度彈性化的校園雲端運算和雲端儲存服務。 本文包含了基本的OpenStack和Ceph配置與部署。我們提出新的虛擬機網路模式(PHA)，讓運算節點可以使用私有IP提供服務，大量節省IP的消耗，並且在正常使用情境下仍然保有HA模式的各種好處；另外使用rgwauthAPI來整合兩個系統上專案和帳號的權限，讓OpenStack可對外提供相容Amazon S3協定的物件儲存服務。

Recent years, cloud computing becomes a hot topic. With the rapid growth of information day by day, the more complex data centers will be set by various institutions. Therefore, how to save data safely and utilize hardware devices efficiently are important issues. Ceph is a new brand of distributed file system. At the design phase, Ceph regards the fault of underlying hardware as normal, and no need to shutoff the whole system when a few machine is scheduled for maintenance offline, thus Ceph can quite ensure the safety of data. Virtualization is a good way to improve the hardware utilization. Traditional virtualization software can easily boot multiple virtual machines on a single server, however, it is hard to manage so many virtual machines, deal with the network connection between virtual machines and different host servers. Hence all the major virtualization companies release corresponding products, such as using VMware ESXi with vCenter to construct the virtualization data center. The other suits, for example Hyper-V role in Microsoft Windows Server and the Citrix XenServer, can quickly provide enterprise private cloud services, but many institutions may flinch from expensive licensing fees and stringent restrictions about hardware specifications. OpenStack is an open source software which is developed by NASA and Rackspace at first. Anyone can use OpenStack to create a free cloud computing services quickly. At last, we establish a private cloud named NCTU CStack system in NCTU ITSC. NCTU CStack is integrated by OpenStack and Ceph storage backend, providing free, highly-flexible campus cloud computing and storage services. This thesis includes the basic configuration and deployment flow about OpenStack and Ceph. We design the new networking mode(PHA), using PHA mode can greatly reduce the consumption of public IP, and it works like HA mode in most scenario. Moreover, we use rgwauthAPI to integrate the authentication between OpenStack and Ceph, and make our system support the object storage service which is Amazon S3 API compatible.