《Oracle虚拟机服务器软件虚拟化在一个64位Linux环境的性能和可扩展性毕业论文外文文献翻译.doc》由会员分享，可在线阅读，更多相关《Oracle虚拟机服务器软件虚拟化在一个64位Linux环境的性能和可扩展性毕业论文外文文献翻译.doc（10页珍藏版）》请在三一办公上搜索。
1、 学号：20090127712013届本科生毕业论文英文参考文献翻译Oracle虚拟机服务器软件虚拟化在一个64位Linux环境的性能和可扩展性（译文） 学院（系）：信息工程专业年级：学生姓名：指导教师：合作指导教师：完成日期：2013年6月Oracle虚拟机服务器软件虚拟化在一个64位Linux环境的性能和可扩展性 Ibidokun Emmanuel Tope, Pavol Zavarsky, Ron Ruhl, Dale Lindskog 加拿大埃德蒙顿康卡迪亚大学阿尔伯塔分校信息系统安全学院摘 要 企业环境采用虚拟化技术带来了巨大的收益，这些还没有它的服务人员问题和异常状况，例如性能调优
2、和不稳定的性能指标，响应虚拟化系统，破坏虚拟化服务器，未配置的虚拟主机平台，或者其他的。研究的焦点是分析Oracle虚拟机服务器虚拟化平台的性能针对那些Bare-metal服务器环境。为了高品质的回报的可扩展性和后续支持也分别分析使用30和50个活跃用户的绩效评估。Swingbench和LMbench是两个基准工具的套装软件被用来测试性能。可扩展性的测试也使用Swingbench。在一个单独的Oracle虚拟机下的Oracle数据库，结果显示使用Swingbench获得了4%和8%的开销对应30和50个活跃用户的绩效评估。相应的75%和87%的性能指标分别对应30和50个活跃用户在两个Orac
3、le虚拟机服务器环境下，表明两个虚拟机可以提高性能可扩展性。我们的结果也表明通过虚拟机服务器测量出的结果有很大的偏差不能被忽略，尽管一些数据是通过LMbench测量。关键词：性能和可扩展性；虚拟化；Oracle虚拟机服务器；Oracle数据库；小红帽；Bare-metal服务器；可利用性第一章 概要目前，很多供销商，比如Intel, AMD, Sun Microsystems, IBM都在发展有着数十亿收益并且不断增长的目标市场的虚拟化产品。虚拟服务器技术有效的将软件和物理设备分离开来。想要一个安全的虚拟服务系统，在应用程序的可用性和进程间的独立性方面都是有改善空间的。虚拟服务器技术同样也遇到
4、了安全方面的挫折，例如虚拟软件的可开发性弱，一些基准测试器件显示出来的性能指标异常，未成形的虚拟主平台和客户操作系统。然而，如果抛开这些安全问题，一旦虚拟环境安全控制松懈，就有产生更大影响的可能。虚拟服务器技术是现代计算机信息处理技术中用来减少计算时间和开销的工具，但是如果被错误运用，它将成为整个环境的安全威胁。这种节省时间与开销的优势促使商业愿意投入更多时间来抓住新的机遇，而不是把精力集中在维护和管理上。虚拟化技术的突出特点就是可扩展性，在虚拟系统中，它在容量，性能和可用性方面有着更大的发展空间。有很多测量基准的工具可以测量不同方面的使用性能。例如，Swingbench可以测试评估每分钟的交
5、易数，LMbench用来测量网络频宽和延迟，Httperf则用来测试网络服务器的性能。而这份研究是评估Oracle数据库在虚拟环境和非虚拟环境下的性能和可扩展性的。我们利用Swingbench和LMbench作为基准测试工具。Swingbench在评估性能方面是很突出的。它的设计目的就是产生负荷压力测试Oracle数据库。而LMbench是通过测试延迟性和频宽给出硬件资源上虚拟技术的影响指数。然后有一些先前的研究【2,9,10,11,12】是评估在32位机的系统上虚拟服务器技术的各方面的。Tolly组织发表过一个研究，是评估在32位的Oracle linux环境下用Oracle虚拟机运行Ora
6、cle数据库企业版的11.1.0.60版本下的性能。我这篇研究的目的在于更新性能标准，以及在64位机上使用不同硬件，软件和配置来扩展这个实验。第二章 相关工作对虚拟机环境的相关性能的评估，有效的工作已经实施。其中一个是由Tolly小组完成评估Oracle虚拟机在32位的环境运行Oracle 11g数据库的性能的白皮书。把solo描述为轻量级虚拟机允许虚拟机以最高的优先级直接运行在硬件上并且通过虚拟化修改后面完成的。它更应该提高传统操作系统的虚拟机性能水平，因此满足高性能应用的需求不惜要唯一的硬件支持。给出一个全面介绍入门管理程序并彻底完成的对Xen虚拟机的性能评估。一个独立的研究通过评估整体性
7、能，安全影响，性能分离和Xen和KVM的可伸缩性来描述了Xen和KVM性能的差别。就虚拟系统的性能评估的基准和工具而论，强调虚拟机监听器是虚拟机系统中作重要的角色它的效率高度影响整个系统的性能。他们也使用LINPACK，LMbench和IOzone测量并分析两个开放的虚拟机监听器Xen和KVM的性能，并且提供一个定量和定性的对两者进行安全的对比。像任何一个新技术，操作和过程不好可能造成很多缺点。至于服务器虚拟化，风险是在虚拟机之间和总体性能指标上缺乏安全性。这项研究打算通过Tolly小组填满64位系统以前的可视化应用。第三章 软件和硬件环境我们这里使用的软件是Oracle虚拟机（Oracle
8、VM server）2.2.1，红帽系统（Red Hat Enterprise Linux (RHEL) ）5.5，）Oracle数据库企业版11g（Oracle Database11g Enterprise Edition Release2），基准测试工具Swingbench2.3，LMbench3-a9。硬件是64位机物理服务器，4核CPU，8G RAM，一个64位虚拟机（1个CPU，3.75G RAM），和两个64位虚拟机（2CPU（每个上面各一个VM），7.5GRAM(每个VM各3.75)）。第四章 实验方法所有的实验评估都是在AMD速龙II（AMD PhenonII），位于多核64位
9、机（3.0GHz,8gb）上进行的。实验环境中用的是同样的电脑为了排除机器条件对实验结果的影响。在单独的的环境下，红帽5.5（Redhat5.5）的操作系统，安装上应用程序Oracle数据库企业版11g（Oracle Database11g Enterprise Edition Release2）。而在虚拟环境中，Oracle虚拟机是虚拟层，在它上面创建两个虚拟机。在每个虚拟机上，都安装上上面单独的环境中一样的软件和程序。另外还要在单独的环境和两个虚拟机环境中都装上Swingbench 和LMbench。测试描述在实验的开始，我们先用Swingbench 2.3.0.422测试并对比单独的服务
10、器所支持的，一个虚拟机上的和两个虚拟机上的每分钟最大事务处理数。把单独的服务器上的测试结果和在虚拟环境下有30,50个活跃用户的使用下的测试结果对比。我们同样也用LMbench3分别测试单独的服务器和Oracle虚拟机环境下的延迟和频宽情况，从而对比他们处理情况和内存性能。我们之所以测试延迟和频宽是因为性能问题通常是由延迟问题，频宽问题或者两者一起引起的。6个实验在反复进行，我们用实验结果的标准差来分析问题。Performance and Scalability Evaluation of Oracle VM Server Software Virtualization in a 64 bit
11、 Linux EnvironmentIbidokun Emmanuel Tope, Pavol Zavarsky, Ron Ruhl, Dale Lindskog Information Systems Security Department Concordia University College of Alberta Edmonton, Canadaeibidokucsa.concordia.ab.ca,pavol.zavarsky,ron.ruhl,dale.lindskogconcordia.ab.caABSTRACTThe growing adoption of virtualiza
12、tion in the enterprise environment has resulted in a couple of hugebenefits, however, this has not been without its attendant problems and anomalies, such as performance tuning and erratic performance metrics, unresponsive virtualized systems,crashed virtualized servers, misconfigured virtual hostin
13、g platforms, amongst others. The focus of this research was the analysis of the performance of the Oracle VM server virtualization platform against that of the bare-metal server environment. The scalability and its support for high volume transactions were also analyzed using 30 and 50 active users
14、for the performance evaluation. Swingbench and LMbench, two open suite benchmark tools were utilized in measuring performance. Scalability was also measured using Swingbench. Evidential results gathered from Swingbench revealed 4% and 8% overhead for 30 and 50 active users respectively in the perfor
15、mance evaluation of Oracle database in a single Oracle VM. Correspondingly, performance metrics of 75% and 87%were obtained with 30 and 50 active users in a dual Oracle VM server, indicating performance scalability improvement with two virtual machines. Our results also revealed significant percenta
16、ges in latency and bandwidth achievement by Oracle VM server which cannot be overlooked, despite some variances in results obtained from LMbench measurement.KEY WORDS:Performance and scalability;Virtualization;Oracle VM server;Oracle Database; RedHat; Bare-metal;server; AvailabilityI. INTRODUCTIONPr
17、esently, many vendors like Intel, AMD, Sun Microsystems, and IBM are developing virtualization products which target markets with revenues in the billions and growing 3. Server virtualization efficiently permits software to be separated from the physical hardware. With a secured server virtualized s
18、ystem, there is improvement on the availability of application and the isolation of processes. Server virtualization also has security setbacks, such as exploitable weaknesses in virtualization software, anomaly performance metrics with some benchmark devices, misconfigured virtual hosting platforms
19、 and guest operating systems 4. However, apart from these diverse security threats, once the security controls for the virtual environment are slipshod, there is a greater possibility for infection. Server virtualization is a tool used to reduce the operational time and cost in todays computing but
20、if done wrong it becomes a security threat to the environment. This saving of time and resources allow businesses to devote more time to new opportunities rather than focusing on maintenance and management 5. A salient feature of virtualization is scalability, whichallows for growth in capacity, per
21、formance and availability of the virtualized system. There are many benchmarking tools designed to measure different performance characteristics. Examples include Swingbench for evaluating transactions per minute 6, LMbench for measuring network bandwidth and latency 7, Httperf for measuring web ser
22、ver performance 8 .This research, evaluates the performance and scalability of the Oracle database both in a virtualized and non virtualized environment. We use Swingbench and LMbench as benchmarking tools.Swingbench benchmarking tool has found prominent in performance evaluation. Swingbench is spec
23、ifically designed to generate load and stress test an Oracle Database. Whereas, LMbench gives an indication of the impact of virtualization on the hardware resources by measuring latency and bandwidth. While there have been previous studies evaluating various aspects of server virtualization on 32-b
24、it architecture system 2, 9, 10, 11, 12. The Tolly group conducted a study for Oracle by evaluating the performance of Oracle VM running Oracle Database 11g Enterprise Edition Release 11.1.0.6.0 in a 32-bit Oracle Linux environment 2. This work aims at updating the performance metrics and extending
25、the experiments with different hardware, software and configuration using 64-bit architecture system 1.II. RELATED WORKSignificant work has been conducted on performance evaluation of various virtualization environments. One such work was the white paper done by the Tolly group which evaluated the p
26、erformance of Oracle VM running Oracle 11g Database application in 32-bit environment 2. Xiang Zhang et al. 11 described solo as a lightweight virtual machine that allows VM to run directly on hardware with the highest precedence and critically alter the overhead done by virtualization. It was furth
27、er stressed that solo enhances VM performance to the level of traditional OS, and thus meets the requirements of the high performance applications without a unique hardware support. Barham et al. 9 gave a comprehensive introduction to the Xen hypervisor and made a thorough performance evaluation of
28、Xen versus VMware. Deshane 12 presented an independent research describing the performance comparison between Xen and KVM, which evaluated the overall performance, security impacts, performance isolation and scalability of Xen and KVM. Regarding the benchmarks and tools for performance evaluation of
29、 virtualized systems, Jianhua Che et al. 10 asserted that virtual machine monitors (VMM) is the main component of virtual machine (VM) system, and its efficiency highly impacts the performance of the entire system. They also measured and analyzed the performance of two open suite virtual machine mon
30、itors-Xen and KVMusing LINPACK, LMbench and IOzone, and provided a quantitative and qualitative security comparison of both virtual machine monitors. Like any new technology, poor implementation and process may create some vulnerability. In the case of server virtualization, the risk would be loss o
31、f security between VMs and overall performance metrics. This research intends to fill the void in equivalent studies for 64-bit architecture system in comparison to the previous version used by the Tolly group 2.III. THE SOFTWARE AND HARDWARE ENVIRONMENTThe software we used in this research are Orac
32、le VM server 2.2.1, the Red Hat Enterprise Linux (RHEL) 5.5, Oracle Database11g Enterprise Edition Release2 application, the benchmark tool Swingbench version 2.3 and LMbench version 3-a9.The hardware: was the machine type-64-bit physical server, 4CPU cores, 8GB RAM, one 64-bit virtual machine with
33、1CPU core 3.75GB RAM size and two 64-bit virtual machines with 2CPU (1 for each VM), 7.5GB (3.75GB for each VM).IV. EXPERIMENTAL METHODOLOGYAll experimental evaluations are performed on the AMD PhenonII, with Quad-core processors 64-bit at 3.0GHz, 8GB. The same physical computer is used in both expe
34、rimentalenvironments to ensure that no machine condition will affect experimental results. In bare-metal environment, Redhat5.5 was the operating system, while Oracle Database 11g Enterprise Edition Release 11.2.0.1.0 was the application used. While in virtualized environment oracle VM was the hyper
35、visor on which two virtual machines were created. On each virtual machines the same software and application was installed in both VMs and bare-metal environment. Swingbench and LMbench were successfully installed on the bare-metal server and on the two virtual machines. A. Test Description In the e
36、xperimental set up, we measured the maximum transactions per minute (tpm) supported by a bare-metal server, versus the transactions served by a single virtual machine and dual virtual machine supporting Oracle VM using Swingbench version 2.3.0.422. The results from the bare-metal server test were co
37、mpared to the results obtained after running the similar test within a virtual environment for 30 and 50 active users respectively. We also used LMbench version 3 in the test to measure latency and bandwidths of both bare-metal server and Oracle VM environment so as to compare the results of their p
38、rocess and memory performance test. We ran the test for latency and bandwidth because performance issues are usually caused by latency problems, bandwidth problems, or some combination of the two. Six iterations of test were run and standard deviation of the results was used for analysis in both tests.
