OpenStack雲端系統效能之研究

Research on OpenStack Cloud System Performance

蘇小芬(Hsiao-Fen Su)

指導教授 ： 柯士文

中原大學/電機資訊學院/資訊工程研究所/碩士(2018年)

https://doi.org/10.6840/cycu201800150

摘要

雲端運算為近年來非常熱門的話題，隨著資訊不斷更新以及開源軟體的發展，要透過網路上所提供的資源建立雲端平台已不是件難事，但基於有無數種架構可建立雲端平台，再加上愈來愈多種儲存環境的需求。因此，找到適用的架構並有效利用也成為非常重要的議題。
在雲端平台上，OpenStack目前已非常普遍部署在雲端系統之上，此為NASA與Rackspace合作研發的開放原始碼雲端運算軟體，能夠操控整個數據中心裡的運算、儲存與網路資源，更提供簡易Web GUI介面讓管理者與使用者更加便於使用。
隨著OpenStack崛起，加上多種儲存環境的需求，Ceph也燃起一陣搭配風潮，Ceph為分散式儲存系統，除了可提供OpenStack所需之區塊式儲存（Block Storage）與物件式儲存（Object Storage）之外，更提供檔案系統（File System）服務，再者Ceph可自成Cluster且支援Scale Out及高容錯機制，也讓Storage在管理上更加簡易。
本文包含基本的OpenStack配置部署外，另與Ceph做選擇搭配，即為OpenStack有搭配Ceph與無搭配Ceph兩種組合，在Ceph部分另外探討其OSD（Object Storage Device）搭配SSD作為Cache是否對效能有影響，因此本實驗環境共分成三組：OpenStack本地儲存（LVM）、OpenStack + Ceph non-Cache及OpenStack + Ceph Cache，透過OpenStack介面於這三組環境上各建立一個Windows Instance（VM）與Ubuntu Instance（VM），再於各VM中安裝硬碟效能讀寫測試工具，利用調整工具中的測試參數進而測試參數是否會影響其效能及較適用於哪種環境上。最後我們將Windows與Ubuntu做為數據對照組，除了探討在三組實驗環境上硬碟讀寫測試之結果外，另將對照組相互比對在不同作業系統平台是否有相同的測試結果。

關鍵字

雲端運算 ； Ceph ； 實體機器 ； OpenStack ； 效能測試

並列摘要

Cloud computing is becoming increasingly important in recent years. With the constant updating of information and the development of open source software, it is no harder to set up a cloud platform through the resources provided on the network. However, there are numerous architectures for building cloud platforms, and the need for more and more storage environments is greater than ever. Therefore, finding the proper architecture and using it effectively has also become a very important issue.
On the cloud platform, OpenStack is now widely deployed on top of the cloud system. NASA is working with Rackspace to develop an open-source cloud computing software that can manipulate computing, storage and network resources throughout the data center. Easy Web GUI interface makes managers and users more user-friendly.
With the rise of OpenStack and the need of multiple storage environments, Ceph has also sparked a pairing with Ceph as a decentralized storage system that can provide both the Block Storage and Object Storage required by OpenStack , It also provides File System services. In addition, Ceph can be clustered to support Scale Out and high fault tolerant mechanisms, making Storage more manageable.
This dissertation contains the basic configuration of OpenStack deployment, the other with the choice of Ceph, that is, with OpenStack with Ceph with no two Ceph, Ceph part of the additional discussion of its OSD (Object Storage Device) as a Cache with SSD performance is effective Therefore, the experimental environment is divided into three groups: OpenStack Local Storage (LVM), OpenStack + Ceph non-Cache and OpenStack + Ceph Cache. Through the OpenStack interface, a Windows Instance (VM) and Ubuntu Instance (VM), and then install the hard disk performance read-write test tools in each VM, and use the test parameters in the adjustment tool to test whether the parameters will affect their performance and which environment is suitable. Finally, we use Windows and Ubuntu as the data control group, in addition to exploring the results of the hard disk read-write test on three experimental environments, and comparing the control groups to each other on the different operating system platforms for the same test result.

並列關鍵字

Cloud computing ； Ceph ； physical machine ； OpenStack ； performance testing

參考文獻

參考文獻

CHEUNG, D. W., NG, V. T., FU, A. W. & YONGJIAN, F. 1996. Efficient mining of association rules in distributed databases. IEEE Transactions on Knowledge and Data Engineering, 8, 911-922.

CHOU, N., LEDESMA, R., TERAGUCHI, Y., BONEH, D. & MITCHELL, J. C. Client-side defense against web-based identity theft. 11th Annual Network and Distributed System Security Symposium, 2004.

COULOURIS, G., DOLLIMORE, J. & KINDBERG, T. 2001. Distributed systems, A distributed system is one in which the failure A distributed computing system consists of multiple.

DIXON, B. E., SIMONAITIS, L., GOLDBERG, H. S., PATERNO, M. D., SCHAEFFER, M., HONGSERMEIER, T., WRIGHT, A. & MIDDLETON, B. A pilot study of distributed knowledge management and clinical decision support in the cloud. Artificial Intelligence in Medicine.