Evaluation of Web Application Servers
Contents
1. Responsible of starting and interrupting the threads a inca ec ae cae tenant are eae ae public void startThreads int choice double iterations switch choice check the choice made One Thread case 1 Process p1 new Process 1 iterations p1 aktivitet start try p1 aktivitet join the main thread this waits for p1 to finish catch Exception e_p1 System out printIn Exception in join break Two Threads case 2 Process p1 new Process 1 iterations 35 Processp 2 new Process 2 iterations p1 aktivitet start p2 aktivitet start try p1 aktivitet join p2 aktivitet join catch Exception e_p2 System out printIn Exception in join 2 threads break Four Threads case 4 Process p1 new Process 1 iterations Process p2 new Process 2 iterations Process p3 new Process 3 iterations Process p4 new Process 4 iterations J p1 aktivitet start a p3 aktivitet start y 0 p2 aktivitet start 0 p4 aktivitet start try p1 aktivitet join p2 aktivitet join p3 aktivitet join p4 aktivitet join catch Exception e_p4 System out printIn Exception in join 4 threads break default System out printin nlllegal argument System out printin Valid argument is 1 2 or 4 V end switch V e2. into apache this to was easily done using the installation wizard for JServ 15 6 1 2 Performance The results from the Apache JServ test on NT can be viewed in table 6 1 As you can see the result differs a lot between the highest and lowest result The lowest value could be the result of a denied request and due to weaknesses in Apache JMeter we were unable to measure them This will affect the calculation of the average and median values and we think that the highest value would be the most accurate value to use in this evaluation HelloWorld html ms HelloWorld class ms StringManip class ms Highest 511 531 9804 Average 96 126 5715 Median 70 110 6469 Lowest 10 10 70 Table 6 1 Result from performance test on Apache JServ As mentioned in chapter 7 a higher degree of accuracy could be obtained by using another performance test application You should however see this results as guidelines and not as statistically correct Nevertheless it can be useful data when comparing between different web application servers 6 1 3 Technologies The different technologies supported by Apache JServ can be viewed in Table 6 2 The lack of support for new hot technologies like CORBA and EJB and the fact that JServ never will support any higher version of Java Servlet API than 2 0 makes this combination less attractive to web application developers Neither Apache nor JServ supports XML but there are
3. Computer Science Per Anders Johansson Magnus Nilsson Evaluation of Web Application Servers Bachelor s Project 2000 10 Evaluation of Web Application Servers Per Anders Johansson Magnus Nilsson O 2000 The author s and Karlstad University This report is submitted in partial fulfillment of the requirements for the Bachelor s degree in Computer Science All material in this report which is not my own work has been identified and no material is included for which a degree has previously been conferred Per Anders Johansson Magnus Nilsson Approved 2000 05 31 Advisor Niklas Nikitin Examiner Stefan Lindskog 111 Abstract A web application server is used to deliver statically or dynamically built web content to cli ents on the World Wide Web This document describes testing of a number of web application servers and the criteria with which you can evaluate them The purpose is to create guidelines that can be used by Ericsson Infotech to evaluate web application servers 1v Credits We would like to recognize our appreciation to the following people Anders Berg our instructor at Ericsson Infotech For everything Niklas Nikitin our supervisor at Karlstad University For guiding us through this project Helena Lindskog For reading our work and for contributing with good ideas on how to evolve it Mikael Johansson For reading our work and for contributing with good ideas on h
4. WebSphere with two built in tools the HTTP server administration tool based on a GUI and the site analysis tool 10 The HTTP server administration tool will help the server administrator configure and manage the HTTP server The site analysis tool can be used to help the server administrator increase the performance of the HTTP server 6 5 Sybase Enterprise Server In this evaluation we will use Sybase Enterprise Server 3 5 Enterprise edition This is a pack age containing several components including Sybase PowerJ 3 5 Sybase PowerBuilder 7 Jaguar CTS and PowerDynamo 6 5 1 Installing Sybase Enterprise Server on Windows NT Installing Sybase Enterprise Server on Windows NT or any other Windows system for that matter is often all about clicking the Next button and so we did During the installation we choosed to include the Jaguar CTS component which is the servlet engine and continued through the installation wizard 1 An HTTP server is equal to a web server both provides the calling client with web pages 12 Microsoft Windows 95 or Microsoft Windows 98 24 6 5 2 Performance Table 6 9 displays the result obtained during our performance test of Sybase Enterprise Server Although this test is done with some accuracy the result is not statistically correct we have only done the performance test once and without statistic methods in mind HelloWorld html ms HelloWorld class ms StringManip cla
5. connection to the other network e A more precise test application which reveals more information about each request e g 1f the web application server is too busy the test application will log this information and not just regard it as a very fast response e Our methods regarding the calculation of test results could be made statistically cor rect e More work could be done to optimize the web application server s performance Both the configurations of the JVM and the web application server could affect the per formance e On some systems the amount of logging could be of importance for the performance We have tried to introduce facts about the different contenders and our intention was to give enough information to be able to compare them on by one We feel however that more could be done to make this evaluation more fair other technologies like SSL and database support could be measured and deeper investigations could lead to more information regard ing tools available for each web application server 28 References m m ANY A aa A W N e kist ah bak A A ee l 8 9 Ben Laurie and Peter Laurie Apache The Definitive Guide O REILLY 1997 Whatis com http www whatis com Whatis com Inc 2000 Jason Hunter and William Crawford JAVA Servlet Programming O REILLY 1998 ServerWatch com http serverwatch internet com Internet com Corp 2000 04 13 Mohammed J Kabir Apache Server Bible IDG Books Wor
6. solu tions to this weakness and one of them is Cocoon from The Apache XML project 8 16 Technology Version Supported CGI Yes CORBA No EJB No JDK 1 2 2 JSP No Servlets 2 0 XML No Table 6 2 Supported technologies in Apache JServ Apache JServ is not JDK dependent this means that you can use any JDK in your web application server runtime environment The ability to execute CGl applications is in cluded in the Apache web server and this might be useful if older web applications are still being used 6 1 4 Price Apache web server and JServ Servlet engine are both freeware and new version can easily be downloaded at The Apache Software Foundation homepage 7 Apart from some legal de tails concerning the Apache JServ product name and under which form redistribution is al lowed anyone is free to use and develop Apache JServ to fit there needs 6 1 5 Support The Apache Software Foundation does not have any support agreement included in their products Instead the user can obtain information about bugs upgrades and more on their homepage 7 The online support contains FAQ s and guides that will help the user solve the most ordinary problems 6 1 6 Tools No tools are included in the Apache JServ configuration The web application environment is established through the use of configuration files e g httpd conf for the Apache web server Since the source code for both Apache and Jserv are availabl
7. technologies mentioned later in this chapter CORBA and EJB can to some extent be used in several of these areas but for sim plicity we will leave them out of this figure HTML XML CGI Technologies CGI implementations JSP Middle tier lt ASP Servlets Java implementations Technologies not End tier lt discussed in this Bachelor s project Figure 5 2 Relationship between technologies 5 3 1 CGI Technologies GET and POST are two ways of delivering data from a web server to an application 5 Sev eral technologies use GET or POST or equivalent methods to obtain the transfer between web 11 server and application In this document we will use the term CGI Technologies to describe this relationship 5 32 CGI implementations CGI was designed to enable a web server to communicate and execute applications outside of the web server A side effect of this is that a web server can receive parameters through an HTML form pass them to a CGI application which returns the result to the server The web server then sends the result as an HTML document back to the client who can use a browser to view it A web server can use two different methods to transfer data to a CGI application the GET method and the POST method 5 Because of the way that CGI works it requires time and a lot of server resources 2 making it less desi
8. 40 2172 4369 The result here is quite good Two threads runs parallel the whole time 4 Processor UNIX 400MHz solstal Solaris_JDK_1 1 7_08a Thread 1 ms Thread 2 ms Thread 3 ms Thread 4 ms Total ms 1 Thread 50000000 Iterations 2130 2137 2 Threads 50000000 Iterations 2132 2130 2137 4 Threads 50000000 Iterations 2130 2127 2126 2128 4263 Here should the threads be split up on all 4 CPU s but the result is the same as on a 2 CPU machine Solaris_JDK_1 2 01 Thread 1 ms Thread 2 ms Thread 3 ms Thread 4 ms Total ms 1 Thread 50000000 Iterations 1764 1777 2 Threads 50000000 Iterations 1987 1879 1998 4 Threads 50000000 Iterations 2426 2200 2440 2333 2539 This is the same test as the latest above but with a newer JDK version As you can see on the result the threads are split up on all 4 CPU s Interesting Table C 1 Performance results from extra task As you can see the JVM version plays a big part in the result The newer JDK_1 2_01 seems to improve the performance a lot compared to the older one 34 C 2 Source code p Thread tester UNIX version Modified 2000 02 25 Author Magnus Nilsson QINXNIM Per Anders Johansson QINXPAJ import java util public class Tt extends Thread public static long thread1 1 public static long thread2 2 public static long thread3 3 public static long thread4 4 Method startThreads Creates instances of inner class Process
9. 6 4 4 Price 6 4 5 Support 6 4 6 Tools 6 5 Sybase Ent rpris Servet ose a e aR E E Cotas estado eee hehe emma ees 24 6 5 1 Installing Sybase Enterprise Server on Windows NT 6 5 2 Performance 6 5 3 Technologies 6 5 4 Price 6 5 5 Support 6 5 6 Tools Y t E esacnes T EO A A A A Installation of Apache and Apache JServ on SolariS ocmonomomsmssssses s es SU B Performance test ai ccsccccssiessccscseccscecdsseccsecdcscsesevescsstdeccseaddesoceacddedecedcdiececseicveccsc esceses cveccsde DL Bi HelloWorld HTML test Par aida 31 B 2 HelloWorld Java Servlet test application oooooonnncccnnoccconoccconancononcnononcnnnonccnnnncnonnnno 31 B 3 StringManip Java Servlet test application ooonococnnococnoncccnnnnnnnonnnononaconanccnnnncnnnnnnos 32 CE EXMTA Tak AA ARO EL RES E ae BE RES E 34 es FOC CONS Bs a Sai a de ds 35 Vil List of Figures Figure 3 1 Three tier application overview nro iia 4 Figure 5 1 Overview of our lab envirOnMent ee eee eeeeeeeeeeseeceeeeeeeeeeaeeceaecneeeseeeeaeens 10 Figure 5 2 Relationship between technologies srta 11 ix List of tables Table 4 1 Evaluation Cri terions siii dani doi ai 6 Table 4 2 Evaluation test Suit siic cc il ied dea elas 7 Table 6 1 Result from performance test on Apache JServ essessesereereesereresrrsresseseresee 16 Table 6 2 Supported technologies in Apache JServ oooooconocccinoccconoccnonancononaconanccnnnna conan 17 Table 6 3 Result f
10. IN showed interest in For further details about the technologies view chapter 5 4 2 3 The price criterion In order for EIN to be able to include a web application server in a project the price of the web application server is essential since it will affect the price of the complete product When looking at the price criterion we will simply check how much the web application server will cost We will only consider the price for the version used in this survey As a unit of measurement we will use the currency USD US Dollar If we have failed to obtain the price in USD and instead found a price in SEK Swedish krona we will transform SEK to USD using an exchange rate of 9 1 Usually we will obtain the price information from the manufacturer homepage or equal 4 2 4 The support criterion When evaluating the support criterion we will look at what kind of support the manufacturer will supply with their product We will check whether upgrades are included in the package and if so for how long time Maybe you will have to pay extra for any upgrade We will also investigate if there are any support number you can call or a mail address where you can get help To proceed with this work we have chosen to use five questions e How long after purchase can the customer get free support e In what ways can the customer receive support Fax phone e mail e At what hours is the support available e Are upgrades free e For how long are upgrades fr
11. LE gt out printin lt HEAD gt out printin lt BODY gt out printin lt H1 gt Hello World lt H1 gt out printin lt BODY gt out println lt HTML gt 31 B 3 StringManip Java Servlet test application id StringManip Test Servlet Magnus Nilsson amp Per Anders Johansson ef RUSS AAS SANS AAS SB AREER EER ERE EE RERRERERR EES REESE ERS RARER ESEREEREASEREES import java io import javax servlet import javax servlet http import java util public class StringManip extends HttpServlet String smallStr new String abcdefghijklmnopqrstuvwxyzaao String bigStr new String long counter 0 long totaltime holds the execution time public void manipulate Date start new Date long total_s_time start getTime get start time while bigStr length lt 100000 manip counter bigStr bigStr smallStr counter Date stop new Date long total_e_time stop getTime get stop time totaltime total_e_time total_s time calculate the difference public void doGet HttpServletRequest request HttpServletResponse respons throws ServletException IOException respons setContentType text html PrintWriter out respons getWriter out printin lt HTML gt out printin lt HEAD gt lt TITLE gt String Manipulator lt TITLE gt lt HEAD gt out printIn lt BODY gt out printin lt BIG gt Performing S
12. age 14 6 Evaluation In this chapter we will evaluate the web application servers Apache JServ Apache Tomcat IBM WebSphere BEA WebLogic Server and Sybase Enterprise application server Together with EIN we came up with the following questions e Performance How will the web application server handle stressful situations like many users at the same time Will it deal with it in a satisfying way e Technologies Does the web application server support any special technologies like Servlets CORBA or CGI e Price What is the cost of buying the web application server e Support Is the web application server supported Does it cost extra Are upgrades included in the price e Tools Does the web application server come with any development environments or any thing else useful like a GUI configuration tool We will try to answer all these questions and summarize it and see if there exists a web application server that is better than all the others 6 1 Apache JServ Apache is one of the most used web servers on the Internet today Combined with JServ Servlet engine Apache works as a web application server In this evaluation we have used Apache web server version 1 3 12 and Apache JServ 1 1 6 1 1 Installing Apache JServ on Windows NT Installing Apache JServ on Windows NT is pretty straightforward First we installed Apache web server by following the installation wizard included The next step was to integrate JServ
13. and information seeking on the Internet More information about our studies is available in chapter 5 1 The first month of this project was used to raise the level of our knowledge As our understanding of web application servers grew we began to look at web application servers available on the market A discussion on how we narrowed the num ber of web application servers down to the five we have decided to use can be viewed in chapter 4 1 Once the contenders were picked we concentrated our work on the evaluation criterions trying to find what was interesting to compare web application servers with Our intention was to produce a list of evaluation criterions that could be used to simplify the evaluation pro cedure For further details on how our web application server evaluation criterion checklist emerged view chapter 4 2 4 1 Choosing the web application servers There are two things you have to consider when you should compare a group of web application servers e The scope how many web application servers should be included e Which web application servers are of interest When we decided how many web application servers to evaluate we tried to balance two opposite factors First the number of web application servers that EIN were interested in and then the amount of time available for this project Considering the given situation we decided that five contending web application servers would be appropriate In the remaining part of cha
14. b sites more application like demanded that web serv ers could display web pages with content that was dynamically built based on output from server side applications An example of server side applications is a calendar application allowing the site user to schedule his or her week The entity that provides the application behavior is often viewed as the middle tier in a three tier application 2 as can be viewed in Figure 3 1 The first tier the front end is often a web based user interface provided by a web server Usually some kind of terminal will re ceive and show the interface to a user but for simplicity we will leave this part out and just view the first three tiers The second tier the middle tier is the business logic that provides the ability to get things done in the application This part of the application often executes on an application server In many enterprise based applications a storage facility is required This is often a database of some sort and it belongs to the third tier the back end of the application European Centre For Nuclear Research The terminal could be a standard PC with a web browser or a cellular phone supporting the WAP technology Business logic Database Figure 3 1 Three tier application overview In some cases the application server is integrated into the web server but it might as well be a standalone server If the application server is integrated into a web server we will
15. connected to Java it supports the latest versions of EJB JSP and Servlets 20 Technology Version Supported CGI Yes CORBA 2 2 2 3 EJB 1 1 JDK 1 2 2 JSP 1 1 Servlets 2 2 XML Yes Table 6 6 Supported technologies in BEA WebLogic Server BEA WebLogic Server also provides XML integration examples that will work with any XML compliant parser and everything is provided for developers to build XML enabled ap plications If CORBA support is wanted you can easily integrate that with BEA WebLogic Enterprise CGI is also supported so those with many CGI applications will not be left out in the cold More information is available at the BEA homepage 11 6 3 4 Price As with many of the web application server developers on the market BEA delivers several different editions of their server We will list some of them below e WebLogic Server Base Edition CPU USD 13 000 e WebLogic Server Cluster Edition CPU USD 20 000 e WebLogic Commerce Server CPU USD 52 000 e WebLogic Personalization Server CPU USD 33 000 e WebLogic Server Developer license license USD 3 300 As you can se the price varies depending on the version you are interested in 6 3 5 Support Once you purchase the BEA WebLogic Server you will have the option to choose what type of support you want Depending on the amount of availability you desire you ll have various packages to choose from The lightest package stretches over office hours 8 hours 5 days we
16. e for everyone tools may already be available or perhaps under development This type of resources usually ends up on the Internet and can be downloaded from there 17 6 2 Apache Tomcat In this chapter we will try to evaluate Apache combined with Tomcat Tomcat is a Servlet engine that is supposed to take over after the JServ Servlet engine and even further down the line be integrated into Apache In this survey we will use Tomcat 3 1 together with Apache 1 3 12 As can be seen in chapter 6 2 3 Tomcat supports newer versions of the Java Servlet API compared to Jserv 6 2 1 Installing Apache Tomcat on Windows NT Installing Apache Tomcat on Windows NT does not differ too much from the installation of Apache JServ First we installed Apache web server by following the installation wizard in cluded The next step was to integrate Tomcat into apache This was done with a little help from the user manual included in the Tomcat distribution 6 2 2 Performance The result from the performance test on Apache Tomcat can be viewed in Table 6 3 As mentioned earlier this result is not statistically correct but it could be used as a guideline dur ing evaluations HelloWorld html ms HelloWorld class ms StringManip class ms Highest 481 1712 15032 Average 99 215 8425 Median 70 140 9153 Lowest 10 20 110 Table 6 3 Result from performance test on Apache Tomcat 6 2 3 Technologies Apache with To
17. eb application servers work and why they are used on the Internet We also felt that Ericsson Infotech provides a good environment to work in and that our project would benefit from the knowledge available there 2 Terminology CGI CORBA EIN EJB FAQ GUI HTML IIS JDK JSDK JSP JVM OMG SSL XML Common Gateway Interface Common Object Request Broker Architecture Ericsson Infotech Enterprise Java Beans Frequently Asked Questions Graphical User Interface Hyper Text Markup Language Internet Information Server Java Development Kit Java Servlet Development Kit Java Server Pages Java Virtual Machine Object Management Group Secure Sockets Layer Extensible Markup Language 3 Background In this chapter we will describe the background to our Bachelor s project It also includes some history about web application servers such as how they got an important place on the Internet and why they are needed We will also describe the main problem and the purpose of our Bachelor s project 3 1 History Our task in this Bachelor s project was to evaluate different web application servers To give a short description on the area of web application servers we have tried to put together a short resume on the subject Ever since the day that Tim Berners Lee built the first web server at CERN 1 descen dants to this early version have been delivering static web content to whom ever connected to the web server The desire to make we
18. ee Our methods in this work are fairly arbitrary We will try to illustrate the support available for each web application server by conducting a discussion based on the questions mentioned above We have obtained the support information through direct contact with representatives from the manufacturers 4 2 5 The tools criterion The list of criterions produced by EIN seems to cover all of the interesting parts of the evaluation work We felt however that something was missing As with all new areas of com puter science web application servers tend to require a certain amount of knowledge in order to profit from all built in features To make it easier to understand and administrate some of the web application servers on the market are equipped with easy to use administration tools We will use this criterion to evaluate if this is the case with any of the web application servers in our survey 5 Background Information In this chapter we will try to give some background information concerning web application servers We will discuss the source of our information and a short description of our lab envi ronment 5 1 Literature When we first began this work we didnt know much about the different technologies avail able to deploy web applications on the Internet Even though our main concern in this bache lor s project was to evaluate which technologies the different web application servers support we thought that a small introduc
19. ek and the heaviest is around the clock support 365 days a year Support is in cluded in the price the first year After that the support fees are paid annually and entitle you to new BEA product upgrades as long as the maintenance and support remains current Cus tomers can receive support via phone fax or e mail 21 6 3 6 Tools The tools included in the test package wasn t that rich but BEA has a package WebGain Studio which can be purchased separately WebGain Studio is an integrated suite of tools to support development of server applications WebGain Studio provides an all round suite that contains all of the products required designing developing debugging and deploying web applications Unfortunately no price information were available for WebGain Studio when we wrote this evaluation 6 4 IBM WebSphere IBM WebSphere can be purchased in three different editions Standard edition Advanced edition and Enterprise edition In this evaluation we will use IBM WebSphere Standard edi tion version 3 02 Some of the differences between the editions will be discussed in chapter 6 4 3 6 4 1 Installing IBM WebSphere on Windows NT The installation was performed in three steps First we installed IBM s own JDK 1 1 7p for Windows NT Then we installed the IBM HTTP Server version 1 3 6 And then finally IBM WebSphere Application Server 3 02 Standard Edition The four other web application servers we tested came with an installat
20. ging make make install After changing back to the Apache source directory we executed make make install Finally we edited the httpd conf file located in the conf directory in the Apache directory There we added this line Include home qinxpaj APP_SERVERS jserv etc jserv conf Finally finished 30 B Performance test B 1 HelloWorld HTML test page lt l gt lt Hello World Test HTML Page gt lt Per Anders Johansson amp Magnus Nilsson gt lt l gt lt HTML gt lt TITLE gt Hello World Test HTML Page lt TITLE gt lt BODY TEXT 000000 BACOLOR FFFFFF gt lt H1 gt Hello World lt H1 gt lt BODY gt lt HTML gt B 2 HelloWorld Java Servlet test application Hello World Test Servlet Per Anders Johansson amp Magnus Nilsson ah FPREESEREREEERER ERS EERE ABEL ARES BEES REESE BERERERE EERE RERERRESE RES import java io import javax servlet import javax servlet http public class HelloWorld extends HttpServlet public void doGet HttpServletRequest request HttpServletResponse respons throws ServletException IOException respons setContentType text html PrintWriter out respons getWriter out printin lt HTML gt out printin lt HEAD gt out printIn lt TITLE gt HelloWorld Test Serviet lt TIT
21. have used five different web application servers and we will now try to summarize the outcome of our evaluation First we have the heavy group containing BEA WebLogic Server and Sybase Enterprise Application Server These servers are capable of handling new technologies and they often come equipped with complete development tools and a large support organization to back them up Our opinion is that these web application servers would suit developers of large web based applications like e commerce web sites In this group we would choose Sybase Enterprise Server as the winner The main reason for this is that Sybase Enterprise Server comes with an easy to understand user interface and good support Also you don t have to edit any configuration files before you can get your Servlets to work On the other hand Sybase Enterprise Server didn t perform as well as BEA WebLogic Server did in the performance test But the overall impression leads Sybase Enter prise Server to victory The other group contains Apache JServ Apache Tomcat and IBM WebSphere Our opin ion based on the impressions gained during this work is that these web application servers though widely spread on the Internet lack the ability to compete with the other two regarding supported technologies and tools available The support for the Java Servlet technology does however give these web application servers a place on the Internet Apache Tomcat and it s predecessor Apache JServ has
22. he most ordinary problems 6 2 6 Tools Apache Tomcat web application server does not come with any fancy development tools but it does include the Ant build tool 7 Ant can be viewed as a Java based equivalent to make in the Unix world 19 6 3 BEA WebLogic Server BEA WebLogic Server has won a lot of awards and they claim to be the market leader with their web application server The BEA WebLogic Server also supports Java and is thought by many to be the best web application server for e commerce In this test we will use version 5 1 6 3 1 Installing BEA WebLogic Server on Windows NT Like most major Windows applications BEA WebLogic Server came with an installation wiz ard By using the default settings it was very easy to install the web application server on our machine 6 3 2 Performance The result from the performance test on BEA WebLogic Server can be viewed in Table 6 5 As mentioned earlier this result is not statistically correct but it could be used as a guide line during evaluations HelloWorld html ms HelloWorld class ms StringManip class ms Highest 921 461 6780 Average 102 74 4184 Median 60 50 4757 Lowest 10 10 100 Table 6 5 Result from performance test on BEA WebLogic Server 6 3 3 Technologies As you can see in Table 6 6 BEA WebLogic Server provides a number of features that today are industry standard components Since BEA WebLogic Server is strongly
23. ion wizard and IBM WebSphere was no exception By fol lowing the wizard it was quite simple to perform the installation 6 4 2 Performance Because IBM WebSphere requires a special JDK version it s not completely fair to com pare the IBM WebSphere performance test with the other web application performance tests Nevertheless the results between the two JDK versions shouldn t differ too much A complete overview of the test result can be viewed in Table 6 7 As with all the other performance tests done during this bachelor s project the statistical value can be discussed We will only consider these values as guidelines 10 TDK 1 1 7 IBM 22 HelloWorld html ms HelloWorld class ms StringManip class ms Highest 671 480 10014 Average 85 123 2717 Median 50 100 2073 Lowest 10 10 60 Table 6 7 Result from performance test on IBM WebSphere 6 4 3 Technologies The Standard edition of IBM WebSphere is pretty modest and some of the techniques sup ported are of older versions IBM WebSphere lacks the ability to run on any JDK and as you can se in our extra task in appendix C the JDK version can have a crucial effect on the per formance In Table 6 8 a complete listing of the supported technologies can be viewed Technology Version Supported CGI Yes CORBA No EJB No JDK 1 1 7p IBM JDK JSP 1 0 Servlets 2 1 XML 1 0 Table 6 8 Supported technologies in IBM WebSphe
24. ldwide Inc 1998 ShopIBM http commerce www ibm com IBM Corp 2000 04 28 The Apache Software Foundation http www apache org The Apache Software Foun dation Org 2000 04 28 The Apache XML Project http xml apache org cocoon The Apache Software Foun dation Org 2000 04 28 The Source For Java Technology http java sun com Sun Microsystems Inc 2000 05 03 10 IBM Corporation http ibm com IBM Corp 2000 05 04 11 BEA Systems http www bea com BEA Systems Inc 2000 05 04 12 Sybase http www sybase com Sybase Inc 2000 05 11 29 A Installation of Apache and Apache JServ on Solaris While standing in the Apache source directory we first executed the following line configure prefix home qinxpaj APP_SERVERS apache make make install Then we edited the httpd conf file located in the conf directory in the Apache directory There we set the variable ServerName To include so shared object support following line was executed configure prefix home qinxpaj APP_SERVERS apache enable rule SHARED_CORE enable module so make install Now we were ready to install JServ While standing in the JServ source directory we executed this line configure prefix home qinxpaj APP_SERVERS jserv with apache src home qinxpaj shared solaris Uncompressed apache_1 3 12 with jdk home opt JDK jdk1 2_01 with JSDK home qinxpaj shared solaris JSDK JSD K2 0Nib jsdk jar disable debug
25. mcat is a solid base for developers with interest for Java Servlet and JavaServer Pages technology Through the concern of Apache this web application server configuration grants access to the CGI technology Even though CGI is on the verge of extinct several web applications still use this technology and they should be able to use Apache Tomcat to overcome the transition Table 6 4 describes in short the technologies supported by Apache Tomcat web application server By older we mean applications based on C C Perl and other 18 Technology Version Supported CGI Yes CORBA No EJB 1 1 JDK 1 2 2 JSP 1 1 Servlets 2 2 XML No Table 6 4 Supported technologies in Apache Tomcat As with Apache JServ XML support can be obtained through the Cocoon 8 add on from the Apache Software Foundation 6 2 4 Price Apache web server and Tomcat Servlet engine are both freeware and new version can easily be downloaded at The Apache Software Foundation homepage 7 Apart from some legal details concerning the Apache Tomcat product name and some limits concerning redistribu tion anyone is free to use and develop Apache Tomcat to fit there needs 6 2 5 Support The Apache Software Foundation does not have any support agreement included in their products Instead the user can obtain information about bugs upgrades and more on their homepage 7 The online support contains FAQ s and guides that will help the user solve t
26. nd startThreads 36 class Process inner class This is where the threads are created Be etre A e o ld Peat Pelee public class Process implements Runnable public Thread aktivitet new Thread this new thread private String text private double iteration private double i constructor public Process String txt double _ iteration text txt iteration _ iteration Method run This is what every thread do public void run try Date start new Date get start time long s_time start getTime for int i 0 i lt iteration i loop baby loop Date finish new Date get the finish time long f_time finish getTime long time f_time s_time how long time did it take if text 1 checks which thread it is thread1 time else if text 2 thread2 time else if text 3 thread3 time else if text 4 thread4 time catch Exception ee System out printIn Exception in class Process method run 37 Main Checks the arguments and starts the program if the arguments are ok A oe senate sence ahaa 7 public static void main String args int choice 1 double iterations 1 String choiceString String iterationString Icheck that number of arguments is correct if args length 2 System out printin nThread Tester 1 0 unix by QINXNIM amp QINXPAJ 2000 System out printIn USAGE java Tt lt a
27. ns As can be seen in Table 4 1 several criterions are measured in qualitative form This means that it is hard to measure the criterion on any given scale To be able to give some kind of answer to the qualitative criterions we will try to conduct a discussion on the outcome of the evaluation 4 2 1 The performance criterion This criterion is an outcome of EIN s desire to obtain information on how fast in millisec onds a web application server can respond to a certain request We will decide this by letting a client invoke three different tasks on a web application server We have decided to use a test suit containing three different applications and we will describe them in Table 4 2 and a com plete view of the source code can be found in appendix B When building our Servlet appli cations we have used Sun JDK 1 2 2 001 The scale reaches from one to five one being the lowest and five the highest grade This information can be obtained by executing java fullversion in the command prompt in Windows NT 4 0 6 Name Purpose HelloWorld html Test the response time for a simple html request HelloWorld class Test the response time for a simple Java Servlet displaying Hel loWorld on the calling client StringManip class Test the response time for a Java Servlet that concatenates string A on to string B until string B is 20000 characters long From the beginning B is empty and A con tains 29 charac
28. ow to evolve it Richard Hellberg Our coach at Ericsson Infotech Contents 4 A A E A 2 Terminol 2 Pea TTL A AA A i are mr ee er 3 A A O 3 32 TS SA E AA A A A AA A E ES 4 DoF O E TO 4 A gt Met A REO 4 1 Choosing the web application servers visi n a a at 5 4 2 Choosing evaluation criterions ist ii dia 6 4 2 1 The performance criterion 4 2 2 The technologies criterion 4 2 3 The price criterion 4 2 4 The support criterion 4 2 5 The tools criterion 5 Background Information cccssccssssccssssccsssscesssscssssscsssscsssssssssssssssssssssssssssesssssssees LO S TEMO UT do E e A A E eN a i 10 322 Tab ENVIO MEA Ia 10 DiS TeChnGlo gies viii si cad d 11 5 3 1 CGI Technologies 5 3 2 CGI implementations 5 3 3 CORBA 5 3 4 Enterprise Java Beans 5 3 5 JDK 5 3 6 JSP 5 3 7 Java Servlets 5 3 8 XML 6 Evaluatio AAA A OR LD Gel A aE a T ES 15 6 1 1 Installing Apache JServ on Windows NT 6 1 2 Performance 6 1 3 Technologies 6 1 4 Price 6 1 5 Support 6 1 6 Tools vi 6 2 Apache Tomica asia 18 6 2 1 Installing Apache Tomcat on Windows NT 6 2 2 Performance 6 2 3 Technologies 6 2 4 Price 6 2 5 Support 6 2 6 Tools 6 3 BEA WebLogic Servet osorno 20 6 3 1 Installing BEA WebLogic Server on Windows NT 6 3 2 Performance 6 3 3 Technologies 6 3 4 Price 6 3 5 Support 6 3 6 Tools 6 4 IBM WebSphere viii dado aces 22 6 4 1 Installing IBM WebSphere on Windows NT 6 4 2 Performance 6 4 3 Technologies
29. pter 4 1 we will discuss how we selected the five contenders EIN had special interest in four web application servers and that left one place open in our survey Since the thoughts behind EIN s interest in some of the contending servers are un known to us we will focus this discussion on how we made our choice At first we knew little about different servers on the market so we began our search on a web site 4 that has spe Apache JServ Apache Tomcat Sybase Enterprise Application Server and IBM WebSphere cialized on evaluating different server software e g web application servers In their opinion BEA WebLogic Server is a five star web application server and one of the best web application servers they have tested With this in mind we thought that BEA WebLogic Server would be a worthy opponent to the web application servers chosen by EIN 4 2 Choosing evaluation criterions This part of chapter 4 will discuss the thoughts behind our evaluation criterions When we first received the specifications for this project made by EIN some evaluation criterions were already mentioned Table 4 1 describes the criterions used in this survey the units in which they are measured and their origin Evaluation criterion Unit of measurement Origin Performance Milliseconds ms EIN Technologies Qualitative EIN Price USD Entity EIN Support Qualitative EIN Tools Qualitative The authors Table 4 1 Evaluation criterio
30. rable as a development platform for web applications To overcome the weaknesses of CGI new techniques have been developed techniques that are less resource consuming A CGI application can be based on different programming languages such as C C or PERL 5 3 3 CORBA CORBA was developed by OMG it is an architecture for creating distributing and managing distributed program objects in a network This means that different programs can communi cate between each other over a network via an interface broker CORBA allows applica tions to communicate with one another no matter where they are located or who has designed them The essential component in CORBA is the Object Request Broker ORB The ORB is the middleware that establishes the client server relationship between objects Using an ORB a client can transparently invoke a method on a server object which can be on the same ma chine or across a network The client does not have to know where the object is located its programming language its operating system or any other system aspects that are not part of an object s interface Doing so the ORB provides interoperability between applications on different machines in distributed environments 5 3 4 Enterprise Java Beans Enterprise Java Beans are reusable components written in Java for Java application develop ment They run in the server side of a client server network relationship An example of this could be a button tha
31. re IBM WebSphere Standard edition comes with support for XML technology this makes it complete as a lightweight application server but with no support for EJB and CORBA de manding web application developers will be better off using the Advanced or Enterprise edi tion 10 6 4 4 Price As with most web application servers IBM provides several distributions of WebSphere ap plication server Potential customers can choose between Standard Advanced or Enterprise edition all with different configuration and function In this evaluation the price criterion will be represented by WebSphere Standard Edition version 3 0 2 distributed on CD ROM As 23 mentioned above several other distributions are available and the price for these can easily be obtained at ShopIBM 6 The price for the distribution we have used is USD 755 and it to was found at ShopIBM 6 6 4 5 Support IBM s support service is quite extensive Several support configurations are available and the price depends on the content of each configuration One support configuration contains an upgrade protection which can be signed for one or two years If the customer choose a sup port configuration with upgrade protection IBM will provide newer versions as they appear Support is available during office hours but for customers with other needs special agree ments can be arranged Support from IBM can be obtained through phone fax or e mail 6 4 6 Tools IBM delivers
32. re the level of support included and the avail ability of the support Some of the plans include new version release which means that newer versions can be obtained through the supplier Users of the Incident Plan will have to purchase an update subscription plan annually in order to receive new releases For customers with really high demands Enterprise Plan Support Option is available This plan will allow the user to personalize the support to fit the need of their projects All of the support plans mentioned above expire after some time and renewal is vital in order to avoid difficulties ob taining support The support is usually available through several forms e g email lists telephone and through the Technical Library CD containing technical documents For multi country instal lations global support agreements can be obtained this will save the customer from having to purchase a separate support agreement in each of the country they are active in 6 5 6 Tools Sybase Enterprise Application Server is a complete suit with development tools and servers Among many other components Sybase Enterprise Application Server contains a tool called PowerJ 12 PowerJ is supposed to help developers create Java programs without problems and in a less time consuming fashion 15 Basic plan Extended plan and Enterprise plan 26 7 Summary It seems that different web application servers have different advantages In this evaluation we
33. read2 ms Thread 3 finished in thread3 ms System out printIn Thread 4 finished in thread4 ms System out println System out println System out printin Total operation time totaltime ms V end else Y end main 39
34. rg1 gt lt arg2 gt System out printin lt arg1 gt is number of threads and n lt arg2 gt is number of iterations in every thread else number of args is ok ftry to extract the arguments try choiceString args 0 iterationString args 1 Double d new Double iterationString Integer i new Integer choiceString iterations d doubleValue choice i intValue catch Exception e catch exception but do nada check if less than zero if choice lt 0 iterations lt 0 System out printin nlllegal argument System out printIn USAGE java Tt lt arg1 gt lt arg2 gt System out printin lt arg1 gt is number of threads 1 2 or 4 n lt arg2 gt is number of iterations in every thread else OK Tt tt new Tt start clocking total time Date totalstart new Date long total_s_time totalstart getTime 38 run threads tt startThreads choice iterations clock the endtime Date totalend new Date long total_e_time totalend getTime long totaltime total_e_time total_s_time if choice 1 System out printin Thread 1 finished in thread1 ms else if choice 2 System out printin Thread 1 finished in thread1 ms System out printin Thread 2 finished in thread2 ms else if choice 4 System out printin Thread 1 finished in thread1 ms Thread 2 finished in th
35. rom performance test on Apache ToMCat ocooncononcnnnncnoncnnoncnoncnannn nncnns 18 Table 6 4 Supported technologies in Apache ToMCat oooonccnnnocnonononnnonnconncnoncnnnnnonnnonnnonns 19 Table 6 5 Result from performance test on BEA WebLogic Server seneese 20 Table 6 6 Supported technologies in BEA WebLogic Server ocooococcnnoccnoonconnconnnonnncnnncnns 21 Table 6 7 Result from performance test on IBM WebSphere oooonocccnoccnoccconccacnnonnnonnnonns 23 Table 6 8 Supported technologies in IBM WebSphere cooooonnoccccoccccconcccnoncconnncncnnnacinnnnnos 23 Table 6 9 Result from performance test on Sybase Enterprise Server cseceeeeees 25 Table 6 10 Supported technologies in Sybase Enterprise Server oooooocccnocccconcccnoncnonnnnnos 25 Table C 1 Performance results from extra taSK oooooonnnnnnnnnonncnnncnconncnonncnnnnonaconnncnncnnnnnos 34 1 Introduction This document is the result of a bachelor s project at Ericsson Infotech in Karlstad Our work began at the end of January 2000 and went on until May 2000 The purpose of our study is to clarify the strengths and weaknesses of different web application servers and thus find out if one is better than the others In this work we have used a number of evaluation criterions and in order to understand the purpose of each evaluation criteria we have chosen to describe them one by one The reason why we chose this bachelor s project is that we wanted to know more about how w
36. s Active Server Pages ASP or Embedded Perl scripts 5 3 7 Java Servlets Servlets are small Java programs small servers that run on web application servers and build web pages dynamically Therefore you can compare it with CGI Building web pages dynamically can be very useful An example is when you use a search engine or put some thing in a shopping cart in a Webshop A Java thread in a single daemon process handles every request that comes in to a Servlet Servlets can also share data among other Servlets which is an advantage when handling with databases Because Servlets are written in Java they are highly portable For example they can be run under Apache Microsoft IIS or almost any major web application server 13 5 3 8 XML XML is like HTML a tag language But unlike HTML where you have a fixed set of tags you define the tags as you go along For example if you like a tag to be named lt BACHE LORSPROJECT gt it s ok Also unlike HTML XML is a valid language that means for every start tag you will need a stop tag or else the parser will complain When you make your XML page you separate content from appearance that means that you have the data in one file and the layout style in another This way authors and designers can work more independ ently XML was derived from SGML Standard Generalized Markup Language and is a meta language A meta language is a language describing another langu
37. see both servers as one entity and call it a web application server A web application server has one major advantage over a standalone application server namely it can provide both a web based user interface and the business logic in one entity In order to deploy applications that can be executed on a web application server develop ers needs a platform Several platforms or technologies are available and we will give a short description on how some of the technologies work in chapter 5 Some of the technologies described are based on the Java technology e g Servlets and JSP and some are based on separate technologies e g CGI 3 2 Task In this bachelors project we will try to answer the following question Is there a web application server that is better than other web application servers 3 3 Purpose When EIN needs a web application server they cant say if one server is better than the other There are also a lot of web application servers on the market which makes it even harder to choose one in specific Each web application server has it s own special trademark and char acteristics Our task is to find out if there exists a web application server that can meet all the require ments that EIN want from such a server 4 Method In the beginning of this bachelor s project our knowledge in this field was inadequate and some studies were needed to understand the given problem Our studies contained both lit erature studies
38. ss ms Highest 1241 501 17055 Average 131 99 7877 Median 90 80 10135 Lowest 10 10 50 Table 6 9 Result from performance test on Sybase Enterprise Server 6 5 3 Technologies Sybase Enterprise Application Server seams to cover it all As can be viewed in Table 4 1 all of the technologies included in this survey are supported Even if we didn find information on which version that is supported in all of the technologies a deeper investigation should give an answer in this matter In Table 6 10 all the supported technologies are listed Technology Version Supported CGI Yes CORBA Yes EJB 1 1 JDK 1 2 2 JSP 1 1 Servlets 232 XML Yes Table 6 10 Supported technologies in Sybase Enterprise Server 6 5 4 Price Sybase Enterprise Application Server is split up in two main packages The package we tested the Development Edition contains among several things a development suite were you can develop your own Java applications The other package Small Business Edition is a little 25 more expensive about USD 3 200 but on the other hand it supports for example database publishing The price for the first package is USD 645 6 5 5 Support Sybase provides support on different levels or plans as they call it The support is divided into 5 plans e Incident Plan e Basic Plan e Extended Plan e Enterprise Plan e Enterprise Plan Support Option The differences between the different plans a
39. t appears on a web page Beans are capsules of code designed for a specific purpose The advantage of Enterprise Java Beans over other standard components is 12 that Beans are independent They are not specific to operating systems or development envi ronments If you create a bean in one development environment you can easily copy and modify it in an another This allows Java Beans to be more flexible in enterprise computing as components are easily shared between developers So instead of upgrading a lot of clients when a component is changed or added you only have to update the server instead of all clients And this way you can distribute program com ponents to clients in a network 5 3 5 JDK A JDK contains all you need to develop compile and run Java applications 9 As with all software newer versions are frequently released and the latest version available at the current time is 1 2 2 5 3 6 JSP JSP or Java Server Pages is a technology invented by Sun Microsystems With Java Server Pages you can create and maintain server side HTML pages These pages can be used both as a kind of dynamic HTML and a replacement for CGI Java Server Pages mixes static HTML with dynamically generated HTML The good side to this is that it s more convenient to write and or modify regular HTML code than to have a lot of println statements that generates HTML which is the case in Servlets You can compare Java Server Pages with Microsoft
40. ters Table 4 2 Evaluation test suit When testing the performance of our web application servers we used the freeware pro gram Apache JMeter Apache JMeter can be used to test overall performance of a web application server For example Apache JMeter can simulate heavy load on a server or network in order to evaluate the performance The output from the program can be presented in a various number of graphical diagrams or you can simply write the output data to a file The performance test will be done in the following way Apache JMeter will simulate one thousand 1000 requests to each server and we will then calculate the highest response time the average response time the median response time and the lowest response time for each web application server The thousand 1000 requests are done by ten 10 different threads and each tread will wait three hundred 300 milliseconds between each call As Table 4 1 implies we will measure the response time in milliseconds ms 4 2 2 The technologies criterion This criterion will be used to evaluate which technologies the web application server s sup port EIN needs this information to decide if a web application server could be used in their projects We have decided to limit the number of technologies in this survey and we will only Apache JMeter can be downloaded at http java apache org jmeter dist 7 Approximately 100 requests by each thread cover the ones E
41. the advantage of being free you can download them and in clude them in your projects without having to pay for them Even though IBM WebSphere have many similarities with Apache JServ and Apache Tomcat one difference is the configuration and management tool included This makes IBM WebSphere more professional in its appearance but bare in mind that IBM WebSphere is a commercial product Trying to summarize this group was not easy but we have decided that IBM WebSphere is the winner Even though some weak spots can be found mostly regarding the old version of the JDK the overall impression is that this web application server is a commercial product with advantages like full scale support fully developed user interface and a fast Servlet en gine 27 We feel that some future extensions could be made to make this work more complete From the beginning of this bachelor s project our intentions was to test the performance in both Windows NT PC and Sun Solaris UNIX environment As we experienced a shortage of time we decided to only run the performance test on Windows NT The method we have used should work fine on any Solaris systems so an extension of our performance test could easily be done Some information on how to install Apache JServ on Solaris can be found in appendix A We also feel that more precise performance tests could be achieved if the following com ponents where made more accurate e A standalone lab environment with no
42. tion to the different technologies would come in handy To make this introduction we have used several books available at EIN 3 5 and some resources on the Internet 2 5 2 Lab environment As you can see in Figure 5 1 our lab environment consists of one server and one client The client 3 is a standard PC running Microsoft Windows NT 4 0 with 64 MB RAM and a Pentium processor working at 133 MHz The server 1 is a PC running Microsoft Windows NT 4 0 with 384 MB RAM and a Pentium II processor working at 266 MHz The computers are connected via a switched 2 network running at 10 Mbit with half duplex transfer rate Entire network Switch 2 PC Client 3 PC Server 1 Figure 5 1 Overview of our lab environment If another platform is desired the server 1 could easily be replaced with another system e g Solaris As mentioned in chapter 4 2 1 we will use the Java based Apache JMeter as our test tool This makes it possible to run the performance test from any Java supporting platform and we can easily replace our workstation 3 to fit our platform needs 5 3 Technologies This part of chapter 5 will discuss the different technologies mentioned in this evaluation Even though several pages or even books could be written about some of the technologies mentioned below our intention is to give a brief information about the different topics that follows Figure 5 2 shows the relationship between the
43. tring Manipulation lt BIG gt manipulate out printin lt BR gt totaltime ms print total time out printin lt BR gt lt BR gt lt BIG gt DONE lt BIG gt out printIn lt BODY gt out printiIn lt HTML gt 32 C Extra Task As a little side task we were asked to develop a program that tested how Java s virtual ma chine handled multiple threads in multiple CPU environment The idea was to start one two or four threads and let them each go through an iterative loop By timing each thread and comparing it to the total time taken we could see if the threads were split up on the CPU s and working parallel The conclusion was that the JDK version played a big part in the result 33 C 1 Results The result from running our test program can be viewed in Table C 1 Multiple threads in multiple CPU environment _ Solaris_JDK_1 1 7_08a Thread 1 ms Thread 2 ms Thread 3 ms Thread 4 ms Total ms 1 Thread 50000000 Iterations 2857 2870 2 Threads 50000000 Iterations 5723 5257 5731 4 Threads 50000000 Iterations 5637 5702 5712 5667 11433 One could think that the total time should be twice as big here because it s a one CPU machine 2 Processor UNIX 400MHz kswas01 Solaris_JDK_1 1 7_08a Thread 1 ms Thread 2 ms Thread 3 ms Thread 4 ms Total ms 1 Thread 50000000 Iterations 2151 2159 2 Threads 50000000 Iterations 2192 2172 2199 4 Threads 50000000 Iterations 2185 2189 21
Download Pdf Manuals
Related Search