TEST REPORT FEBRUARY 2008 Performance of three Intel-based SMB servers using WebBench, LoadSim, and SysBench Executive summary
Intel Corporation (Intel) commissioned Principled Technologies (PT) to measure the performance while running multiple benchmarks simultaneously on three dual- and quad-core small and medium-sized business (SMB) focused servers: Intel Pentium Dual-core processor E2160based server (1.80 GHz, 1 MB L2 cache, 800 MHz bus, 65nm) Quad-Core Intel Xeon processor X3220based server (2.40 GHz, 8 MB L2 cache, 1066 MHz bus, 65nm) Quad-Core Intel Xeon processor X3350based server (2.66 GHz, 12 MB L2 cache, 1333 MHz bus, 45nm)

KEY FINDINGS

The 45nm Quad-Core Intel Xeon processor X3350-based server delivered an average of 161.3 percent better performance across the three benchmarks than did the 65nm Intel Pentium Dual-core processor E2160-based server (see Figure 1) at an additional cost of about $389 (see Appendix B). The 45nm Quad-Core Intel Xeon processor X3350-based server delivered an average of 34.2 percent better performance across the three benchmarks than did the 65nm QuadCore Intel Xeon processor X3220-based server (see Figure 1) at an additional cost of about $131 (see Appendix B).
We ran benchmarks that simulated three aspects of a typical small business server workload: Web, email, and database. We used WebBench* to simulate Web traffic, Microsoft* Exchange Server* 2003 Load Simulator* (LoadSim) to simulate email activity, and SysBench to simulate the database activity. The Workload section provides more information on each of the benchmarks. Figure 1 presents averaged results for the three servers on the three benchmarks relative to the lowestperforming server, the Intel Pentium Dual-core processor E2160-based server. For each benchmark, we assigned a value of 100 percent to that servers results and then calculated the percentage performance improvement of both the Quad-Core Intel Xeon processor X3220-based server and the Quad-Core Intel Xeon processor X3350based server. This approach makes each data point a comparative number, with higher numbers indicating better performance. Finally, for each server, we took the percentage performance improvement over the Intel Pentium Dual-core processor E2160-based server for the three benchmarks and averaged them.
Averaged peformance relative to the Intel Pentium Dual-core processor E2160-based server 3.0

Normalized performance

2.5 2.0 1.5 1.0 0.5 0.0
Dual-Core Intel Pentium processor E2160based server Quad-Core Intel Xeon processor X3220-based server Quad-Core Intel Xeon processor X3350-based server
As Figure 1 illustrates, the Quad-Core Intel Xeon processor X3350-based server delivered an average of 161.3 percent better performance across the three benchmarks than did the Intel Pentium Dual-core processor E2160-based server. The Quad-Core Intel Xeon processor X3350-based server delivered an average of 34.2 percent better performance across the three benchmarks than did the Quad-Core Intel Xeon processor X3220-based server.

Processor

Figure 1: Averaged performance of the three test servers across the three benchmarks relative to the Intel Pentium Dual-core processor E2160-based server. Higher numbers are better.
Normalized performance per dollar results 1.8
1.6 1.4 1.2 1.0 0.8 0.6 0.4 0.2 0.0
Dual-Core Intel Pentium processor E2160-based server Quad-Core Intel Xeon processor X3220-based server Quad-Core Intel Xeon processor X3350-based server
For each of the three servers, we took the averaged relative performance results in Figure 1 and divided them by the price (see Appendix B). For ease of comparison, we then normalized those results to that of the Intel Pentium Dual-core processor E2160-based server. As Figure 2 illustrates, the Quad-Core Intel Xeon processor X3350based server produced a 65 percent higher performance per dollar result than the Intel Pentium Dual-core processor E2160-based server.
The Quad-Core Intel Xeon processor X3350-based server produced a 16 percent higher performance per dollar result than the Quad-Core Intel Xeon processor X3220-based server. Of the entry servers we tested, the 45nm Intel Xeon processor-based server offers the best SMB value.

Figure 2: Normalized performance per dollar result of the test servers across the three benchmarks. Higher numbers are better.

Workload

WebBench WebBench 5.0 (128-bit US version) is an industry-standard benchmark for Web server software and hardware. It uses PC clients to send Web requests to a server under test. It generates performance results by incrementally increasing the number of clients making HTTP 1.0 GET requests to the Web server; the result is a curve showing the servers performance under increasing load. The peak of that curve represents the peak throughput of the server. WebBench reports both the total number of requests per second the server handled and the servers total throughput in bytes per second. To be certain that we found the true peak performance in our testing, we verified that the servers processor utilization reached or was extremely close to 100 percent during the test. LoadSim Microsoft Exchange Server 2003 Load Simulator is a benchmark that measures a servers performance handling email. LoadSim uses the MAPI Messaging Benchmark 3 (MMB-3) to simulate MAPI email clients. The simulated clients issue requests to an Exchange server to generate an email load. LoadSim measures both the servers response time and the number of messages it delivers. LoadSim is available for free download from www.microsoft.com/downloads/details.aspx?FamilyId=92EB2EDC-3433-47CA-A5F80483C7DDEA85&displaylang=en. SysBench The developers at SourceForge.net (www.sourceforge.net) created SysBench to test various aspects of the performance of servers running database systems. The benchmarks original target was the MySQL database system (see www.mysql.com). Intel created a version of SysBench 0.4.0 that would work with Microsoft SQL Server 2005 Enterprise Edition (www.microsoft.com/sql/editions/enterprise/default.mspx). We ran that version in our test. SysBench created a 100,000-row SQL database and executed a batch of online transaction processing (OLTP) transactions against that data.
Principled Technologies, Inc.: Performance of three Intel-based SMB servers using WebBench, LoadSim, and SysBench
Normalized performance per dollar

Test results

For testing, we staggered the starting times of the benchmarks to allow each server to ramp up slowly. We began the benchmarks in the following order: WebBench, Microsoft LoadSim, and SysBench. Figure 3 shows a timeline for the benchmark runs. To obtain the final results, we removed the first and last parts of each benchmark run, and computed the results based on the 30 minutes of peak performance during the SysBench run.
Multiple benchmark timeline 1.5 hours total time WebBench

00:45 - 01:15

LoadSim

SysBench

Figure 3: The timeline we followed when starting the benchmarks for the multiple-benchmark runs on all servers. The shaded areas from 00:45 to 01:15 represent the period that we measured performance.
Figure 4 presents results for the three servers on the three benchmarks relative to the lowest-performing server, the Intel Pentium Dual-core processor E2160-based server. For each benchmark, we assigned a value of 100 percent to that servers results and then calculated the percentage performance improvement of both the QuadCore Intel Xeon processor X3220-based server and the Quad-Core Intel Xeon processor X3350-based server. This approach makes each data point a comparative number, with higher numbers indicating better performance. Finally, for each server, we took the percentage performance improvement over the Intel Pentium Dual-core processor E2160-based server for the three benchmarks and averaged them. As Figure 4 shows, the Quad-Core Intel Xeon processor X3350-based server achieved more Web requests per second and more database transactions per second than the other two servers. It achieved these wins with less processor utilization.

Server

Percentage CPU utilization
Web requests per second (relative to Intel Pentium Dual-core processor E2160-based server ) 100% 323% 386%
Email response time (relative to Intel Pentium Dual-core processor E2160-based server ) 100% 122% 111%
Database transactions per second (relative to Intel Pentium Dual-core processor E2160-based server ) 100% 145% 287%
Average of Web, email, and database scores (relative to Intel Pentium Dual-core processor E2160-based server ) 100% 197% 261%
Intel Pentium Dualcore processor E2160-based server Quad-Core Intel Xeon processor X3220-based server Quad-Core Intel Xeon processor X3350-based server

99.4 99.2 90.0

Figure 4: Performance relative to the Intel Pentium Dual-core processor E2160-based server for the three servers on the three benchmarks. Higher numbers are better.
WebBench results We ran WebBenchs default ecommerce API test suite, which generates both secure and non-secure static and dynamic HTTP 1.0 GET requests. While running the ecommerce suite, the clients must negotiate to a secure Web server port using the Secure Socket Layer (SSL) protocol. A default WebBench test suite incrementally increases the number of clients making the HTTP 1.0 GET requests to the Web server. As the workload increases the number of clients, the Web servers processor utilization also increases, until the processor in the Web server is saturated with work. Each workload point with a fixed number of clients is a WebBench mix. The ecommerce API test suite begins with a mix that involves one client; the next mix has four clients, and each subsequent mix increases the number of clients by four to a total of 60 clients. We modified the test suite so it would run 20 clients with 10 engines per client for the entire test. This allowed us to keep a constant Web load on the server. We performed the standard 16 mixes for testing. A WebBench run reports the total requests per second a server can perform and the total throughput, in bytes per second, that the server delivered. WebBench reports these results for each mix. To obtain the results below we averaged the requests per second and throughput from the mixes that ran during the peak performance of the benchmark run as illustrated in Figure 3. Figure 5 shows the WebBench peak results in requests per second and in throughput (bytes per second) for the three test servers. Each result below is the median of three runs.

Server Intel Pentium Dual-core processor E2160-based server Quad-Core Intel Xeon processor X3220-based server Quad-Core Intel Xeon processor X3350-based server
Throughput (bytes per second) 30,546,722.7 98,636,204.2 117,905,941.5
Requests per second (raw) 7,534.8 24,355.5 29,089.9
Requests per second (relative to Intel Pentium Dual-core processor E2160-based server) 100% 323% 386%
Figure 5: Median WebBench results for the three servers. Higher numbers are better.
LoadSim results LoadSim results show the response time each for each server, the time in milliseconds it took to complete a request. Since the object of the server is to handle as many requests as possible, lower scores, which show a server able to handle more requests in a given time, are better. Generally the most important point is the median 95th percentile, which means 95 percent of the time the server handled request at that level. People usually run LoadSim tests on servers with large disk storage systems, i.e., a large number of total disk drives. Since our testing focused on small business servers, which do not typically have a large number of disks, the scaling between the servers was not as good as the other benchmarks. The systems were mostly bottlenecked waiting on the disk, prohibiting a large amount of scaling. This bottleneck created an inverse relationship between SysBench and LoadSim results: as SysBench scores increased, LoadSim scores dropped and vice versa. For testing we used 100 Microsoft Exchange mailboxes; all of the servers could handle the load while performing other tasks. However, the Quad-Core Intel Xeon processor X3350-based server handled the load better than the Intel Pentium Dual-core processor E2160-based server while the Quad-Core Intel Xeon processor X3220-based server handled the load better than both other servers. The Quad-Core Intel Xeon processor X3220-based servers faster response time is most likely due to the aforementioned disk bottleneck as the Quad-Core Intel Xeon processor X3350-based server wins SysBench by a large margin. Median 50th percentile 98 Median 95th percentile 185 Median mean 112 Median 95th percentile (relative to Intel Pentium Dual-core processor E2160-based server) 100% 122% 111%

Figure 6: Median LoadSim results for the three servers. Lower numbers are better.
SysBench results For testing, we created a script so each system would perform as many transactions as possible in a given time. The script processed 10,000 transactions as quickly as possible and then started over again. The output shows the completion time in seconds for each group of 10,000 transactions. We calculated the transactions per second by using the following formula: total transactions = transactions per second time in seconds Figure 7 shows the median SysBench results and processor utilization for the three test servers. Worth noting, the Quad-Core Intel Xeon processor X3350-based server and Quad-Core Intel Xeon processor X3220-based server performed more transactions per second with less processor utilization that the Intel Pentium Dual-core processor E2160-based server. This means both systems had resources available to perform additional tasks. Transactions per second (relative to Intel Pentium Dual-core processor E2160-based server) 100% 145% 287%
Server Intel Pentium Dual-core processor E2160based server Quad-Core Intel Xeon processor X3220based server Quad-Core Intel Xeon processor X3350based server
Completion time (seconds)

Transactions completed

Transactions per second

30:00 30:00 30:00

133,701 194,094 383,328

74.3 107.8 213.0

Figure 7: Median SysBench results for the three servers. Higher numbers are better.

Test methodology

Figure 8 summarizes some key aspects of the configurations of the three server systems; Appendix A provides detailed configuration information. Intel Pentium Dualcore processor E2160-based server 1.8 GHz 800 MHz Dual S3000AH Intel 3000 Chipset Kingston KVR667D2E5K2/1G 2 x 512 MB SDRAM Samsung SP2004C Intel PRO/1000 PM and Intel PRO/1000 MT Quad-Core Intel Xeon processor X3220based server 2.4 GHz 1,066 MHz Quad S3000AH Intel 3000 Chipset Kingston KVR667D2E5/1G 2 x 1024 MB SDRAM Samsung SP2004C Intel PRO/1000 PM and Intel PRO/1000 MT Quad-Core Intel Xeon processor X3350-based server 2.66 GHz 1,333 MHz Quad S3210SH Intel 3210 Chipset Kingston KVR800D2E5K2/2G 2 x 1024 MB SDRAM Samsung SP2004C Intel PRO/1000 MT
Server Processor frequency (GHz) System bus Dual/Quad-Core processors Motherboard Chipset RAM Hard drive NICs

Figure 8: Key aspects of the server configurations.
Intel configured and provided all three servers.
We began our testing by installing a fresh copy of Microsoft* Windows* Server 2003 R2, Enterprise Edition Service Pack 2 on each server. We followed this process for each installation: 1. 2. 3. 4. 5. 6. Assign a computer name of Server. For the licensing mode, use the default setting of five concurrent connections. Enter a password for the administrator log on. Select Eastern Time Zone. Use typical settings for the Network installation. Use Testbed for the workgroup.
We applied the following updates from the Microsoft Windows Update site: Windows Server 2003 Cumulative Security Update for Internet Explorer 7 for Windows Server 2003 (KB937143) Windows Update Windows Server 2003 Security Update for Windows Server 2003 (KB936021) Windows Update Windows Server 2003 Security Update for Windows Server 2003 (KB938127) Windows Update Windows Server 2003 Cumulative Security Update for Internet Explorer 6 for Windows Server 2003 (KB937143) Windows Update Windows Server 2003 Security Update for Windows Server 2003 (KB921503) Windows Update Windows Server 2003 Security Update for Windows Server 2003 (KB936782) Windows Update Windows Server 2003 Windows Malicious Software Removal Tool - August 2007 (KB890830) Windows Update Windows Server 2003 Security Update for Windows Server 2003 (KB926122) Windows Update Windows Server 2003 Security Update for Microsoft.NET Framework, Version 1.1 Service Pack 1 (KB933854) Windows Update Windows Server 2003 Security Update for Windows Media Player 6.4 (KB925398) Windows Update Windows Server 2003 Update for Windows Server 2003 (KB936357) Windows Update Windows Server 2003 Security Update for Windows Server 2003 (KB935839) Windows Update Windows Server 2003 Security Update for Windows Server 2003 (KB935840) Windows Update Windows Server 2003 Cumulative Security Update for Outlook Express for Windows Server 2003 (KB929123) Windows Update Windows Server 2003 Security Update for Windows Server 2003 (KB924667) Windows Update Windows Server 2003 Update for Windows Server 2003 (KB927891) Windows Update Windows Server 2003 Security Update for Windows Server 2003 (KB932168) Windows Update Windows Server 2003 Windows Internet Explorer 7.0 for Windows Server 2003 Windows Update Windows Server 2003 Security Update for Windows Server 2003 (KB931784) Windows Update Windows Server 2003 Security Update for Windows Server 2003 (KB930178) Windows Update Windows Server 2003 Security Update for Windows Server 2003 (KB925902) Windows Update Windows Server 2003 Update for Windows Server 2003 Service Pack 2 (KB931836)
We updated the X3220 and X3350-based systems as little as possible to make them comparable with the E2160based system from August. However, we had to update the BIOS and chipset drivers to make the two new systems work properly. We also had to update the RAID controller drivers in order to allow the Windows installation to properly identify the RAID array on the X3220 and X3350-based systems. We did not, however, update Windows or any other drivers past the August date of the low-end system.

Installing and configuring the mail server
To set up this server, we had to install several additional software components. The following subsections detail the necessary installation processes. Before following these instructions, make sure to statically assign an IP address for the server under test, and then plug that port into another system or router. This allows the server to correctly make the domain server. Installing the domain controller 1. Select Start Administrative Tools Manage Your Server. 2. Click Add or remove a role.
Principled Technologies, Inc.: Performance of three Intel-based SMB servers using WebBench, LoadSim, and SysBench 7
3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20.
At the Configure Your Server window, click Next. At the Configuration and Options screen, select Custom Configuration, and click Next. At the Server Role screen, select Domain Controller, and click Next. At the Summary of Selections screen, click Next. At the Welcome pop-up window, click Next. At the Operating System Compatibility screen, click Next. For the Domain Controller Type, leave the default option (Domain controller for a new domain), and click Next. At the Create New Domain screen, leave the default (Domain in a new forest), and click Next. At the Install or Configure DNS window, select No, and click Next. At the New Domain Name screen, type a domain name (loadsim.com), and click Next. At the NetBIOS Domain Name screen, enter a NETBIOS domain name (loadsim), and click Next. Click Next through the following screens: Database and Log Folders, Shared System Volume, and Permissions. Doing so accepts the default options for each of those screens. At the Directory Services Restore Mode Administrative Password screen, enter a password (password), and click Next. When the installation software prompts you to do so, insert the OS CD, and click OK. At the Completing the Active Directory Installation Wizard screen, click Finish. When the system prompts you to reboot the server, click Restart Now. After the server reboots, log in to the system. At the This Server is Now a Domain Controller window, click Finish.
Installing Internet Information Services 6.0 1. Select Start Control Panel Add or Remove Programs. 2. Click Add/Remove Windows Components. 3. Select Application Servers, and click Details. 4. Click ASP.NET, and make sure a check appears in the check box. 5. Select Internet Information Services (IIS), and click Details. 6. Click NNTP Services and SMTP Services, make sure a check appears in both check boxes, and click OK. 7. Click OK to close the Application Server window. 8. At the Windows Components Wizard, click Next to begin the installation. 9. When the system prompts you to do so, insert the OS CD, and click OK. 10. At the Completing the Windows Components Wizard window, click Finish. 11. Close the Add or Remove Programs window. Installing Exchange Server 2003 1. Insert the Microsoft Exchange Server 2003 CD. The CD should automatically launch the installation software. 2. Select Exchange Server Enterprise Edition. 3. Run the Setup application. 4. Click the Exchange Deployment Tools link. 5. Click the Deploy the first Exchange 2003 Server link. 6. Click the New Exchange 2003 Installation link. 7. Click the Run ForestPrep now link. a. When the compatibility issue notice appears, click Continue, and check Dont display this message again. b. At the Welcome screen, click Next. c. Accept the End User License Agreement, and click Next. d. At Component Selection Screen, keep the defaults, and click Next. e. Leave the default account information, and click Next. Installation should start. (This process may take a while to complete.) f. When the installation completes, click Finish. 8. Return to the Exchange Server Deployment Tools screen, and click the Run DomainPrep now link. a. At the Welcome screen, click Next. b. Accept the End User License Agreement, and click Next. c. At Component Selection Screen, keep the defaults, and click Next. d. At the insecure domain notice, click OK.

Deploying WebBench data WebBench includes data that must reside on the server and that the Web server must use. We loaded that data and set the Web server to use it with the following procedure: 1. Copy the file wbtree.exe from the WebBench CD to the wwwroot directory on the server under test. (The wbtree.exe file is on the WebBench CD at \wb50\workload). 2. On the server, execute the wbtree.exe file. This program copies the WebBench workload to the server. 3. In the wwwroot folder on the server, create a new folder with the name CGI-BIN. 4. Copy the file simisapi.dll to the CGI-BIN folder. 5. Click Start Programs Administrative Tools Computer Management to open the management console. 6. Go to Services and Application Internet Information Services Default Web Site CGI-BIN, and rightclick to bring up a menu. Select Properties. 7. Go to the Directory Tab, and Set Execute Permissions to Scripts and Executables. 8. When the IIS WWW Configuration popup window displays, click Yes. 9. Close the Computer Management window. Configuring Internet Information Services (IIS) We configured the Windows Internet Information Services Web server as follows: 1. Open Computer Management.
Principled Technologies, Inc.: Performance of three Intel-based SMB servers using WebBench, LoadSim, and SysBench 9
2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24.
Go to Services and Application Internet Information Services. Right-click to bring up a menu, and select Properties. Click MIME Types. In the MIME Types window, click New. In the Extension field, type *. In the MIME type field, type application/octet-stream, and click OK. Click OK to exit the MIME Types window. Click OK to exit the Internet Information Services (IIS) Manager Properties window. In the Computer Management window, go to Web Sites Default Web Site under Internet Information Services, right-click it to bring up a menu, and select Properties. Select the Home Directory tab. Check the following boxes: Script source access, Read, Write, and Directory browsing. Uncheck the following boxes: Log visits and Index this resource. In the Execute permissions field, select Scripts and Executables. Click Apply. Select the Web Site tab. Uncheck Enable Logging. Click OK to close the Default Web Site Properties window. In the Computer Management window, go to the Web Service Extensions item under Internet Information Services. In the right panel, highlight All Unknown ISAPI Extensions, and click Allow. Acknowledge the IIS Manager warning that appears by clicking Yes. In the right panel, highlight All Unknown CGI Extensions, and click Allow. Acknowledge the IIS Manager warning that appears by clicking Yes. Close the Computer Management window.

Installing certificate services Because WebBench includes tests that involve security, we installed Windows Certificate Services as follows: 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. Click Start Control Panel Add/Remove Programs. Select Add/Remove Windows Components. Check the box beside Certificate Services. Accept the notice that the servers name cannot be changed after you install this service. Click Next. Select Stand Alone Root CA, and click Next. On the CA Identifying Information window, enter the server name in the Common name for this CA field, and click Next. Leave the default Certificate Database Settings as they are, and click Next. At the prompt asking if you want to enable Active Server Pages (ASP) in IIS, click Yes. At the prompt that explains that Internet information Services will be temporarily stopped, click Yes. Click Next. Click Finish. Click Close.
Setting up Internet Explorer We had to set up Internet Explorer on the server as follows to be able to get the necessary certificates: 1. 2. 3. 4. 5. Click the Internet Explorer icon on the Quick Launch toolbar. Select the Tools Internet Options menu item. On the Connections tab, click LAN Settings. Check Automatically detect settings, and click OK. Click OK to exit the Internet Options window.
Requesting the server certificate We had to set up Internet Explorer on the server as follows to be able to get the necessary certificates: 1. Open Internet Explorer and enter the following website, http://server/certsrv. This should bring up a page entitled Microsoft Certificate Services server. 2. Click Request a Certificate. 3. Click Advanced Certificate Request. 4. Click Create and submit a certificate request to this CA. 5. Fill out the appropriate Identifying Information details on the Advanced Certificate Request. Name: Performance Company: PT Type of certificate needed: Server Authentication Certificate Key Options: o CSP: Microsoft Base Cryptographic Provider v1.0 o Usage: Exchange and Signature select both o Key size: 512 o Check Automatic key container name o Check Store certificate in the local computer certificate store Additional Options: o Hash Algorithm: SHA-1 Otherwise use default 6. Click Submit. 7. Acknowledge the Potential Scripting Violation by clicking Yes. A page informing you that there is a Certificate Pending should appear. Issuing the server certificate Next we had to issue a server certificate as follows: 1. 2. 3. 4. 5. Select Start Programs Administrative Tools Certification Authority. Click the servers name. Select Pending Requests. Right-click Pending Requests, select All Tasks, and select Issue. Close Certification Authority.

Installing and configuring the database server
Installing Microsoft Internet Information Server 1. Select Start Administrative Tools Manage Your Server. 2. Select Add or remove a role. 3. At the preliminary steps screen, click Next. 4. At the Server Role screen, select Application Server, and click Next. 5. At the Summary of Selections screen, click Next. 6. When the installation software prompts you to do so, insert the OS CD, and click OK. 7. When the installation completes, click Finish. 8. Close the Manage Your Server window. Installing Microsoft SQL Server 2005 1. Insert the SQL Server 2005 CD into the server. 2. Click the Server Components link under the Install section to begin the installation. 3. Accept the license agreement, and click Next. 4. Click Next to install the prerequisites. 5. After the Prerequisites complete installing screen, click Next. 6. At the Welcome screen, click Next. 7. At the System Configuration Check, assuming everything has a status of success, click Next. 8. Enter the registration information, and click Next. 9. At the Components to Install screen, select SQL Server Database Services, then click Advanced. 10. Instruct the software to install both the Database Services and Client Components by clicking the dropdown arrow on both and selecting Entire feature will be installed on local hard drive. 11. Click Next to start the installation. 12. For Instance Name, leave the default option, and click Next.
Principled Technologies, Inc.: Performance of three Intel-based SMB servers using WebBench, LoadSim, and SysBench 12

13. 14. 15. 16. 17. 18.

At the Service Account screen, select Use the built-in system account (Local account), and click Next. Keep the default option for Authentication Mode, and click Next. For Collation Settings, leave the defaults, and click Next. Accept the default Error and Usage Report Settings, and click Next. At the Ready to Install screen, click Install to begin the installation. When the installation software prompts you to do so, insert Microsoft SQL Server 2005 disk 2, and click OK. 19. When the setup process completes, click Next. 20. Click Finish to complete the installation.
Creating the test database and generating data 1. Select Start Microsoft SQL Server 2005 SQL Server Management Studio. 2. In the Connect to Server window, use the drop-down menu next to Server name to select Browse for more. 3. In the Browse for Server window, select the Network Servers tab, and wait for it to retrieve data. 4. When Database Engine appears, expand the menu, select Server, and click OK. 5. At the Connect to Server window, the server name should be present. 6. Click Connect to connect to the server. 7. Right-click the Database folder in the left window, and select New Database from the pop-up menu. 8. In the Database name field, type sbtest, and click OK. 9. Open a command prompt by selecting Start Run, type cmd, and click OK. 10. In the command prompt, type cd to change to C:\. 11. To prepare the database, type vsysbench.exe --test=oltp --oltp-table-size=100000 prepare, and press Enter. 12. When the command finishes, the C:\ prompt appears. 13. Type exit, and press Enter to close the command prompt. 14. In SQL Server 2005 Management Studio, stop the database service by right-clicking server in the left window pane, and then select Stop from the drop-down menu. 15. Using Windows Explorer, create the folder C:\backup. 16. With Windows Explorer, copy all sbtest files from C:\Program File\Microsoft SQL Server\MSSQL.1\MSSQL\Data\ to c:\sysbench_backup. (You should see two files: sbtest.mdf and sbtest_log.ldf.) 17. Close Windows Explorer once the copy completes. 18. Restart the database service using SQL Server 2005 Management Studio by right-clicking server and selecting start.

Installing the mail test client
Each mail server requires one physical test client system. We used the following process to set up each such system. 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. Install Microsoft Windows Server 2003 R2 Enterprise Edition. Use mailclient as the system name. Install Windows Server 2003 Service Pack 2 with the default settings. Open the Network Connections window by right-clicking My Network Places and selecting Properties. Right-click the network connection, and click Properties. Highlight TCP/IP, and select Properties. Select Use the following IP address, and enter the correct information. Click OK to close the two Properties windows. Right-click My Computer, and select Properties. Click the Computer Name tab. Click Change. Select Domain in the member of section, and type the domain test.com into the text box. In the pop-up window, enter administrator for user name and password for password, and click OK. At the Welcome to the domain screen, click OK. When the warning that you must restart dialog appears, click OK. Click OK to close the System Properties window. Click Yes to the pop-up message to restart the computer.

17. 18. 19. 20.

Login to the domain by using administrator and password. Install Microsoft Outlook 2003 with the default settings. (Outlook 2003 doesnt require any configuration.) Install LoadSim 2003 by double-clicking the loadsim.msi installation file. Accept all defaults. Open LoadSim 2003, and make sure no error messages appear.
Configuring LoadSim 2003 1. Start LoadSim by clicking Start Menu All Programs Exchange LoadSim LoadSim 2003. 2. From the LoadSim menu bar, select Configuration Topology Properties. 3. Click the Servers tab. 4. In the left pane, click the + next to the name of this server. 5. In the left pane, select the First Storage Group. 6. In the right pane, set the number of users to 100. 7. Select the Security and Credentials tab. 8. Leave the Credentials choice set to Login to users using the account LoadSim is running on. 9. In the password portion of the tab, type Password1, and click OK. 10. From the LoadSim menu bar, select File Save As. 11. Save a backup copy of the simulation file as C:\loadsim.sim. 12. From the LoadSim menu bar, select Run Create Topology, and wait until the Creating Users and Creating DLs tasks complete. 13. From the LoadSim menu bar, select Run Initialize Test. 14. When a dialog box appears with the question Do you want to initialize Public Folders from this client machine?, click Yes. 15. The run is complete when the bottom left corner of the LoadSim window shows Status: Idle. This task typically takes 2 to 3 hours to complete. Back up the mail database Once youve set up LoadSim and created its initial mail database, you need to back up that database so you can have clean copies for each test. Use this process to back up that database. 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. Select Start All Programs Microsoft Exchange System Manager in the mail server. In the left pane, expand Servers until First Storage Group appears highlighted. In the right pane, right-click Mailbox Store, and select Dismount Store from the menu. Click Yes on the Do you want to continue? pop-up message. When youve dismounted the Mailbox Store, a red arrow appears in the window pane. Right-click Public Folder Store, and select Dismount Store from the menu. Click Yes on the Do you want to continue? pop-up message. The red arrow appears when youve dismounted the Public Folder Store. Using Windows Explorer, create a new folder C:\backup. With Windows Explorer, copy all files from C:\Program Files\Exchsrvr\mdbdata to C:\backup. This process may take several minutes. Close Windows Explorer. In Exchange System Manager, right-click Mailbox Store, and select Mount Store from the menu. Click OK on the pop-up message, The store was successfully mounted. Right-click Public Folder Store, and select Mount Store from the menu. Click OK on the pop-up message, The store was successfully mounted. Close Exchange System Manager.

Network test bed configuration
To generate the workload, we used a network with 22 client PCs. Each PC contained an Intel Pentium 4 3.0GHz with HT Technology processor, 512MB of system memory, a 40GB hard drive, and a Gigabit Ethernet network adapter. We configured each client with Windows XP Professional with Service Pack 2 and all critical updates available as of June 7, 2007. We split 20 of the clients into two segments, or subnets, of 10 clients each. We connected each segment to the server under test via one NETGEAR* GS724T Gigabit Smart Switch*. To balance the load on the server and to prevent a network throughput bottleneck, we connected each segment to a separate network port. We used the
Principled Technologies, Inc.: Performance of three Intel-based SMB servers using WebBench, LoadSim, and SysBench 14
remaining two clients for the WebBench controller and LoadSim client. Figure 9 illustrates the test bed configuration.
Figure 9: Illustration of the test network we used to generate the WebBench workload.
Appendix A Test system configuration information
This appendix provides detailed configuration information about each of the test server systems, which we list in alphabetical order.

Servers

General Number of processor packages Number of cores per processor package Number of hardware threads per core System power management policy CPU Vendor Name Stepping Socket type Core frequency (GHz) Front-side bus frequency (MHz) L1 cache L2 cache Platform Vendor and model number Motherboard model number Motherboard chipset Motherboard revision number Motherboard serial number BIOS name and version BIOS settings Chipset INF driver Memory module(s) Vendor and model number Type Speed (MHz)
65nm Intel Pentium Dual-core processor E2160-based server 1 Always on Intel Intel Pentium ESocket 775 LGA 1.8 GHz 800 MHz 32 KB + 32 KB (per core) 1 MB (shared) Intel S3000AH Intel AZAX71300522 Intel Corporation S3000.86B.02.00.0043.061820070639 Default Intel 8.1.1.1010 Kingston KVR667D2E5K2/1G PC2-5300 DDR2 SDRAM ECC unbuffered 667 MHz
65nm Quad-Core Intel Xeon processor X3220-based server 1 Always on Intel Intel Xeon XSocket 775 LGA 2.4 GHz 1,066 MHz 32 KB + 32 KB (per core) 8 MB (2 x 4 MB) Intel S3000AH Intel AZAX71200495 Intel Corporation S3000.86B.02.00.0043.061820070639 Default Intel 8.1.1.1010 Kingston KVR667D2E5/1G PC2-5300 DDR2 SDRAM ECC unbuffered 667 MHz

3. Making Computer Detect the New Device
Run the CMOS(BIOS) set-up program to detect the new hard disk drive.

SAMSUNG

HDD INSTALLATION GUIDE
1 Go into CMOS screen during the booting of the computer. 2 Make the computer detect the hard disk.
This product is a precision designed mechanism and is very sensitive to shock and electrostatic discharge (ESD). When storing, please avoid high temperature and humidity and use the origirnal ESD protection package in which the product was shipped at the time of purchase. Please handle the product carefully. Damaged or detached labels or any other components may directly cause product malfunction and void warranty. 1. Configuring your Hard Disk Drive (Master, Slave)
In case drive capacity is 32GB or more, please refer to bellow diagram.

General Pin Setting

For descriptive details of the program, refer to the computers users manual
If your computer supports IDE HDD Auto Detection, the hard disk drive will automatically be detected. If your computer does not support auto detect, go into the User Define Mode and by referring written information on the product, set the correct settings for Cylinder, Head, and Sector and if some Bios must be, set the HDD Mode to LBA. Save the changed details and exit.
Setting as Master All other Series Pin Setting

(AB + FH)

Setting as Slave

(BD + FH)

Cable Select

(BD + EF)

4. Partitioning and Formatting the Drive
Refer to your Operating System manual or system manual for partitioning and formatting.
Pin Setting (32 GB Clip Pin Setting)
5. Samsung World Wide Web Site
Additional information can be obtained at www.samsunghdd.com. Cable Select

(CD + EF)

Setting as Master

(AB + CD)

(CD + FH)
2. Installing Your Hard Disk Drive
Note: Turn off the system power !
Cover opening mechanism and the location of the hard disk drive may vary from one computer to another.
SAMSUNG ELECTRONICS CO., LTD. INFORMATION MEDIA BUSINESS STORAGE SYSTEM DIVISION HEAD OFFICE: 94-1, IM SOO-DONG, KUMI-CITY KYUNG BUK, KOREA 730-350 FAX : 82-54-479-5567 This device complies with Part 15 of the FCC Rules. Operation is subject to the following two conditions : (1) this device may not cause harmful interference, and (2) this device must accept any interference received, including interference that may cause undesired operation. ' Samsung Electronics Co., Ltd. Printed in Korea. BF68-00099B Rev.03
Turn off the computer, unplug the external cables, and open computer cover. Mount the hard disk drive in the designated place in the computer with screws provided with the product. Connect 40 pin interface cable and power cable. Make sure that the direction of the cables are correct and matches the shape of the receptacles when connecting cables. Incorrect cable connection may damage the hard disk drive.

Number 1Pin