压力测试及性能数据分析

1
1- 3
Abstract 概要 The objective of this test was to assess and analyze the scalability and performance of a web application that used PHP scripts running on Sun Solaris 7 using Apache web server and Oracle database. The specific goals included the following: Address possible causes of performance bottlenecks. Determine how fast users can expect web pages to be returned to them. 这篇文章的目标是评估和分析一个使用PHP脚本语言运行在Sun Solaris7 使用Apache web服务器和Oracle 数据库的web应用程序的可测性和性能。本文明确的目标包括如下： 网址可能引起性能瓶颈。 检测多快用户能期待网页能够被返回（打开）。 We used Microsoft‘s Web Application Stress Tool, WAS, to meet these objectives. 我们使用Microsoft的Web应用压力测试工具（WAS）来完成这个目标。 This paper presents the process and results of our findings. We tested the web application on a Netra T1 1440, Solaris 7 running Apache 1.3.12. A Sun Ultra 10 running Solaris 7 and Oracle 8.1.7 was used as the database server. We collected stress data and other statistics, including requests per second (RPS) and total time to last byte (TTLB). The maximum RPS data infers the number of concurrent users that a server platform can support. The TTLB data shows how fast the results are presented to the client by the server. 这文章呈现了过程和我们所发现的结果。我们测试web应用程序在Netra T1 1440、运行 了Apache1.3.12的Solaris 7上。一个运行在Solaris7 的Sun Ultra10和Oracle8.1.7 用作数据库服务器。我们收集压力数据和其他的统计数据，包括每秒请求数（RPS）以及 接受到最后一个字节的平均时间（TTLB）。从每秒最多请求数数据推断出一个服务器能支 持的并发用户数。 The test environment is described in Section 1.1, and the test results are presented in Section 2.2. Section 2.3 explains the data analyses. Section 1.1描述了测试环境，在Section2.2中给出了测试结果。Section2.3解释了数据 分析。
1-1
WEBSTAR
Web Application Stress Test and Data Analysis压力测试及性能数据 分析
April 15,2002 by Khalid Anwar and Arif Saleem
1-2
Table of Contents 1 WEB APPLICATION STRESS TESTING ................................................ 1-3 1.1 Test environment ..............................................................1-3 1.1.1 Hardware Platform Configurations................................................1-3 1.2 Web Stress Test Plan .......................................................1-4 1.2.1 Web Application Stress Tool .....................................................1-4 1.3 Typical Trial Report ..........................................................1-8 2. Data Analysis........................................................... 1-10 2.1 Maximum Requests per Second .................................................... 1-10 2.2 Maximum Average Load on the server ......................... 1-12 2.3 Response Time ............................................................. 1-13 2.3.1 Time To Last Byte (TTLB) ............................................... 1-13 2.3.2 Time To First Byte (TTLB) ............................................... 1-13 2.4 Socket Connects ........................................................... 1-13 2.5 Failure ........................................................................ 1-14 2.6 Bytes Sent Rate/ Bytes Receive Rate ................................ 1-14 3. SUMMARY OF REPORT ..................................................................... 1-15 LIST OF FIGURES Figure 2-1. Application Settings ...................................................................1-5 Figure 2-2. Typical Trial Summary Report......................................................1-9 Figure 2-3. Number of Threads and RPS (POST method, No BW throttling)...... 1-10 Figure 2-4. Number of Threads and RPS (GET method, No BW throttling)........ 1-11 Figure 2-5.Number of Threads and RPS (56K BW throttling) ........................... 1-11 LIST OF TABLES Table 2-1. Client Test Configurations ...........................................................T局域网、最大带宽为100Mbits/s的实验室环境下执行。 PHP脚本语言运行在NetraT1 1440的服务器， Solaris操作系统并且Apache1.3web服务 器上。Netra有一个single 440MHz UltraSparc CPU and 1G的内存。Oracle8.1.7数 据库运行在Sun Ultra10的服务器上，同样运行着一个single 440MHz UltraSparc CPU and 512MB 内存的Solaris7系统。（翻译的有问题） 1.1.2 User Accounts and Client Machine A maximum of 250 user accounts were established on the WAS client machines to simulate access to the servers. 一个最大数量为250用户账号在WAS的客户机上被建立来模拟对服务器的存取。 The WAS tool client was installed on a Microsoft Windows 2000 workstation with a 533-Mhz processor and 187 MB of RAM. To stress a powerful server, either one powerful client or several less powerful clients are set up to provide the load. WAS工具客户端被建立在一个533-Mhz处理机和187MB内存的Windows2000的工作站上。为了 压力测试一个强有力的服务器，我们建立了一个强有力的客户机或者几个弱一点的客户机 来提供负载。 We determined the maximum number of requests per second that the Web server can handle, the number of concurrent users when failures start, bytes sent and received rates, the maximum average load on the server; we discussed server resource bottlenecks and presented system response times. Section 2.1 describes how the tests were performed, Section 2.2 presents a typical test result, and Section 2.3 presents the conclusions of our Web application stress tests. 我们测定web服务器能够处理的最大数量的每秒请求数，当失效发生时的并发用户数，字 节发送和接受率，在服务器上的最大平均负载；我们论述服务器资源瓶颈和给出系统回复 时间。Section2.1描述了测试如何被执行，Section2.2给出了一个典型的测试结果，以及 Section2.3给出了我们web应用程序压力测试的结果。 1.2 Web Stress Test Plan（Web压力测试计划） 1.2.1 Web Application Stress Tool（Web应用压力测试用具） The Microsoft Web Application Stress (WAS) tool is designed to simulate multiple browsers requesting pages from a Web site. This tool can realistically simulate many requests with relatively few client machines. Microsoft‘s tutorial on the WAS tool includes information on installation, scripting, performance counters, settings, and reporting. It also provides general guidelines for using WAS and describes common Web testing problems. The tutorial is located at the WAS tool site (). The following paragraphs describe our setup to perform these tests. 微软的Web应用压力测试工具 （WAS） 是被设计来模拟对一个Web地址的大量的浏览器请 求。这工具能够实际地模拟从少许的相关客户机发出大量的请求。微软的WAS工具的用户 指南包括安装、脚本、性能计数器、设计和报告的信息。指南也提供一般的使用WAS以及 描述平常的Web测试问题的指导方针。这指南在WAS网址 （）有下载。接下来的段落描述了我们执行这些测试 的设置。 1.2.1.1 Application Settings（应用软件设置）
1 Web Application Stress Testing （Web应用程序压力测试）
1.1 Test environment（测试环境） 1.1.1 Hardware Platform Configurations（硬件平台配置） The test was performed in a lab environment, on an isolated 100BaseT local area network (LAN), giving a maximum bandwidth of 100Mbits/s. The PHP scripts were run on the Netra T1 1440 Server, with Solaris 7 operating System and Apache 1.3web server. The Netra had a single 440MHz UltraSparc CPU and 1GB RAM. The Oracle 8.1.7 database was run on a Sun Ultra 10 server, also running Solaris 7,with a single 440MHz UltraSparc CPU and 512MB RAM.