Understanding and Using NetBench(TM) 2.10 LICENSE AGREEMENT FOR ZIFF-DAVIS' NETBENCH(TM) 2.1 READ THIS AGREEMENT CAREFULLY BEFORE USING THE SOFTWARE EMBODIED IN THE NETBENCH(TM) 2.1 DISKETTE (OR, IF DOWNLOADED, IN THE DOWNLOADED FILE(S)). Embodied in the NetBench 2.1 diskette ("diskette") (or, if downloaded, in the downloaded file(s)) is the NetBench 2.1 computer program and related documentation (the "Software"). Ziff-Davis Publishing Company, L.P., having a place of business at One Park Avenue, New York, New York 10016 ("Ziff") is the licensor under this Agreement and you are the licensee. By using the Software, in whole or in part, you agree to be bound by the terms of this Agreement. If you do not agree to the terms of this Agreement, promptly return the Software (or, if downloaded, delete the Software) to the Ziff-Davis Benchmark Operation at One Copley Parkway, Suite 510, Morrisville, North Carolina 27560. Title to the Software and all copyrights, trade secrets and other proprietary rights therein are owned by Ziff. All rights therein, except those expressly granted to you in this Agreement, are reserved by Ziff. 1. Limited License This Agreement grants you only limited rights to use the Software. Ziff grants you a non-exclusive, non-transferable license to use the Software on a file server networked with multiple PC computers for the sole purpose of conducting benchmark tests to measure the performance of computer hardware and operating system configurations. You have the right to make a single copy of the Software for archival purposes and the right to transfer a copy of the Software across a network only to the PC computers attached to the network. You may not publish or distribute benchmark test results obtained by you from your use of the Software without prior written permission from Ziff-Davis. For such permission, contact the Ziff-Davis Benchmark Operation at the above address. This Agreement and your rights hereunder shall automatically terminate if you fail to comply with any provision of this Agreement. Upon such termination, you agree to cease all use of the Software, to delete the Software and to destroy all copies of the diskette and other materials contained in this package in your possession or under your control, or, if downloaded, to destroy any and all copies of the Software in your possession or under your control. 2. Additional Restrictions A. You shall not (and shall not permit other persons or entities to) rent, lease, sell, sublicense, assign, or otherwise transfer the Software or this Agreement. Any attempt to do so shall be void and of no effect. B. You shall not (and shall not permit other persons or entities to) reverse engineer, decompile, disassemble, merge, modify, include in other software or translate the Software, or use the Software for any commercial purposes, except for the publication or distribution of test results with Ziff's prior written permission, as provided above. C. You shall not (and shall not permit other persons or entities to) remove or obscure Ziff's copyright, trademark or other proprietary notices or legends from any of the materials contained in this package or downloaded. 3. Limited Warranty and Limited Liability THE SOFTWARE IS PROVIDED "AS IS" WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESS OR IMPLIED, INCLUDING, WITHOUT LIMITATION, ANY WARRANTY OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. THE ENTIRE RISK AS TO THE RESULTS AND PERFORMANCE OF THE SOFTWARE IS ASSUMED BY YOU, AND ZIFF ASSUMES NO RESPONSIBILITY FOR THE ACCURACY OR APPLICATION OF OR ERRORS OR OMISSIONS IN THE SOFTWARE. IN NO EVENT SHALL ZIFF BE LIABLE FOR ANY DIRECT, INDIRECT, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE USE OR INABILITY TO USE THE SOFTWARE, EVEN IF ZIFF HAS BEEN ADVISED OF THE LIKELIHOOD OF SUCH DAMAGES OCCURRING. ZIFF SHALL NOT BE LIABLE FOR ANY LOSS, DAMAGES OR COSTS, ARISING OUT OF, BUT NOT LIMITED TO, LOST PROFITS OR REVENUE, LOSS OF USE OF THE SOFTWARE, LOSS OF DATA OR EQUIPMENT, THE COSTS OF RECOVERING SOFTWARE, DATA OR EQUIPMENT, THE COST OF SUBSTITUTE SOFTWARE OR DATA, CLAIMS BY THIRD PARTIES, OR OTHER SIMILAR COSTS. THE ONLY WARRANTY MADE BY ZIFF IS THAT THE ORIGINAL PHYSICAL MEDIA IN WHICH THE SOFTWARE IS EMBODIED AND WHICH IS DISTRIBUTED BY ZIFF SHALL BE FREE OF DEFECTS IN MATERIALS AND WORKMANSHIP. ZIFF'S ENTIRE LIABILITY AND THE USER'S EXCLUSIVE REMEDY SHALL BE LIMITED TO THE REPLACEMENT OF THE ORIGINAL PHYSICAL MEDIA IF DEFECTIVE. THE WARRANTIES AND REMEDIES SET FORTH HEREIN ARE EXCLUSIVE AND IN LIEU OF ALL OTHERS, ORAL OR WRITTEN, EXPRESS OR IMPLIED. NO ZIFF AGENT OR EMPLOYEE, OR THIRD PARTY, IS AUTHORIZED TO MAKE ANY MODIFICATION OR ADDITION TO THIS WARRANTY. SOME STATES DO NOT ALLOW EXCLUSION OR LIMITATION OF IMPLIED WARRANTIES OR LIMITATION OF LIABILITY FOR INCIDENTAL OR CONSEQUENTIAL DAMAGES; SO THE ABOVE LIMITATIONS OR EXCLUSIONS MAY NOT APPLY TO YOU. 4. U.S. Government Restricted Rights. The Software is licensed subject to RESTRICTED RIGHTS. Use, duplication or disclosure by the Government or any person or entity acting on its behalf is subject to restrictions as set forth in subdivision (c)(1)(ii) of the Rights in Technical Data and Computer Software Clause at DFARS (48 CFR 252.227-7013) for DoD contracts, in paragraphs (c)(1) and (2) of the Commercial Computer Software-Restricted Rights clause in the FAR (48 CFR 52.227-19) for civilian agencies, on in other comparable agency clauses. The contractor/manufacturer is the Ziff-Davis Benchmark Operation, One Copley Parkway, Suite 510, Morrisville, North Carolina 27560. 5. General Provisions Nothing in this Agreement constitutes a waiver of Ziff's rights under U.S copyright laws or any other Federal, state, local or foreign law. You are responsible for installation, management, and operation of the Software. This Agreement shall be construed, interpreted and governed under New York law. If any provision of this Agreement shall be held by a court of competent jurisdiction to be illegal, invalid or unenforceable, the remaining provisions shall remain in full force and effect. Trademarks NetBench(TM), WinBench(TM), WinMark(TM), Winstone(TM), PC Bench(TM), DOSMark(TM), and MacBench(TM) are trademarks and WinBench(R) is a registered trademark of Ziff-Davis Publishing Company, L.P. Microsoft(R) and MS-DOS(R) are U.S. registered trademarks and Windows(TM) is a trademark of Microsoft Corporation. NetWare and Novell are registered trademarks of Novell, Inc Corp. Preface This manual tells you how to install, set up, and run NetBench(TM). NetBench is one of several benchmark programs developed by the Ziff-Davis Benchmark Operation to help you test your PCs, servers, and Macintoshes. For information on getting copies of other benchmarks or additional copies of NetBench, see Appendix F. Notes about NetBench WARNING The user is cautioned against running these tests on a "live" production network. NetBench is designed to exercise the network in various ways. Running NetBench should be done in isolation when other user activity on the network is non-existent. Results would be meaningless if the network were active with user tasks. Please note the following before you run NetBench: 1. You must read and agree to the license information in the front of this manual before running NetBench. The same licensing information appears on your PC's screen when you first run NetBench and on-line in the README file. If you do not agree to the licensing information, remove NetBench entirely from your system and return it and all accompanying materials (including any documentation) to ZDBOp at the following address: The Ziff-Davis Benchmark Operation One Copley Parkway, Suite 510 Morrisville, NC 27560 2. System Requirements The minimum configuration required to run NetBench is one computer which acts as a file server and another computer which runs the NetBench test station software. This test station accesses the file server through a logical drive. NetBench Diskette Contents ABOUT TXT Information about NetBench BATCH TXT Sample Batch command file LICENSE TXT License text agreed to by user (also located in Document) LICFILE TXT Created by License procedure NETBENCH EXE Actual NetBench executable file PARMS CFG Used in Batch example PARMS1 CFG Used in Batch example README.TXT Readme file referred to within NetBench USERINFO TXT Licensed to text (user and organization) Created after the first time a program is licensed to a user. ZD_NB210 PCX Startup Graphics Screen ZDLPARMS CFG Essential file required for NetBench to execute NETBENCH TXT ASCII readable Text version of User Guide NETBENCH RTF Rich Text Format - User Guide NETBENCH DOC Word for Windows version 2.0a - User Guide Table of Contents Chapter 1: NetBench Overview 1 NetBench is a tool 1 NetBench acts like an application 2 Model of Network Application Layers 2 The results reflect the system configuration 3 Using NetBench 4 Chapter 2: Installing NetBench 7 Installing on a PC 7 Installing NetBench on a Network Server 9 What's Next? 11 Example 2A (below). 11 An example of the contents of Autoexec.bat : 11 An example of the contents of config.sys: 12 An example of WS.ID contained in the root directory of C:\ 12 STARTUP.NCF 12 AUTOEXEC.NCF 12 Chapter 3: Running NetBench 13 NetBench Displays 13 Ziff-Davis' NetBench Screensaver 14 Waiting Display 15 Activating the Test Station Coordinator 16 Basic Navigation - Your choice: Mouse or Keypress 17 Navigation Keys 17 NetBench Menus and sub-menus 18 File Menu 19 Rules for Batch files 20 Valid Syntax 20 Valid Batch Examples 21 Performance Menu 23 Options Menu 24 Set Name 24 General 24 NIC Options 26 I/O 27 Chapter 4: Network Interface Card (NIC) Throughput Test 31 Factors that influence the test 32 Parameters 32 Special Rules 33 Guidelines 33 Testing the server NIC 33 Testing the test station NIC 34 Setting-up and Running the Test 34 Evaluation Scenarios 36 Chapter 5: I/O Throughput Test 39 Mechanics 39 Objective 39 Factors that influence the test 41 Parameters 41 Guideline 43 Setting-up and Running the Test 43 Evaluation Scenarios 45 Appendix A: Data File and Statistics File Parameters 47 Appendix B: Temporary Files 51 Appendix C: ZDLPARMS.CFG 53 Appendix D: Data File Output Example 55 Appendix E: Statistics File Output Example 67 Appendix F: Technical Support 75 Contacting ZDBOp 75 Benchmark Request Form Glossary Chapter 1 NetBench Overview This guide describes what NetBench is, how it is installed, the basics of setting up and running each of the NetBench tests, and some scenarios of how each NetBench test has been used during pre-release testing to measure network file server performance. It includes some cautions and suggestions, but it cannot anticipate all possible uses (proper or improper) for NetBench. You are responsible for understanding the objective of your test and how you use NetBench. NetBench is a tool NetBench is a tool that helps determine relative file server performance. In order to do this, two distinct tests were developed: Network Interface Card, and I/O Throughput. The purpose of each test is to isolate, exercise, and measure throughput over time at the respective system location: the network interface card, or the file server disk I/O subsystem. These tests are influenced by the operating system used by the network, memory, memory cache, CPU types and speeds, server disk speeds, and Network Interface Cards -- the same items that influence the performance of a user application on a network. Although the fundamental operation of a NetBench test does not change, modifications to test parameters such as test duration and file size, as well as modifications to the network system under examination produce results which can be comparatively analyzed and evaluated. Proper interpretation of the results produced allow the user to determine what type of network configuration is most effective for a simulated network load. NetBench acts like an application NetBench was designed to "act like an application" (see Figure 1) and is not reliant on any specific network operating system. In other words, "special system calls" were avoided so that the specific network operating system used would be transparent to the testing process. The goal of the software is to test File Services (the capacity of the file server for file I/O activity), and to test the capacity of the network through the network interface card. The tests allow the user to specify one of two locations in the system to generate a simulated "bottleneck" of activity. This helps the user determine the network configuration best suited for a particular application environment. The term "bottleneck" is deliberately chosen. It represents the maximum number of times an "activity" is performed in a user-specified time frame. The maximum number of "activities" performed depends on the specific hardware chosen for the network and is also influenced by the number of test stations configured for the test. Finally, it is important to keep in mind that because each test station performs the maximum number of test iterations possible during the test period, an actual application load of the same network configuration would likely perform better. NetBench tests worst-case scenarios. Typical applications on a production network tend to be less resource demanding than the "bottleneck" tests performed under NetBench. Model of Network Application Layers The following provides you with an example of the different layers that are in a network application. A. Application Layer running on a test station NetBench viewpoint -------------------------------------------------- B. Network Redirector (redirected drive) C. Network Adapter Driver (protocol e.g. IPX, IP, DECNET) D. Network Interface Card (Workstation adapter card) E. Wire F. Network Interface Card (Network server adapter card ) G. Server Figure 1. Representation of a Simulated Network Application In this example, everything that appears below the dashed line (i.e., B through G) represents information hidden from NetBench. The results reflect the system configuration The results are representative of the specific configuration used (hardware, software, and NetBench parameters). The results may be compared properly when one variable at a time is modified in the configuration. Careful thought must be taken when choosing the variable to change. Some modifications may have little or no effect on the network. Understanding which items make sense to vary really depends on how well you (the NetBench user) understand the network configuration that you are using. Using NetBench The following sequence is suggested to help you get started using NetBench: 1) Install NetBench. In Chapter 2, the section, Installing NetBench on a PC, provides instructions for installing NetBench on a single-user PC system should you desire to become familiar with the tool before installing it on a network. Network testing cannot be accomplished using the PC installation procedure. The section, Installing NetBench on a Network Server, provides instructions for installing NetBench on a server drive for actual network testing. The Installing NetBench on a PC, section should not harm a network as long as NetBench is installed on the local C: drive, whereas the network installation procedure will seriously impair the network (and compromise its functioning) if the network were active with user tasks. At the same time, the values returned as a result of the test will be meaningless. 2) Decide on a test. Determine which NetBench test to run: NIC Throughput or I/O Throughput. 3) Read about the test. Read the corresponding section in this guide about the test. 4) Configure test network. Make any test-specific adjustments to the test network. 5) Understand the NetBench menus. Use the Running NetBench section to learn about using the menus to configure NetBench parameters for the test. 6) Set any NetBench parameters. Use the menus to set the NetBench parameters. 7) Run the test. Chapter 2 Installing NetBench This chapter talks about how you can install and start using NetBench. It includes information on installing NetBench on a PC and on a network server. Installing on a PC This section provides instructions for installing NetBench on a PC in order become familiar with the tool. Network testing cannot be accomplished using the PC installation procedure. Load the NetBench files in an empty directory on the PC. To do this: 1. Create the directory on the PC. Any directory name will work. The drive must be local with respect to the PC. In this case we will assume the C: drive. md fakenet In this example we will use FAKENET. NOTE: This directory must be empty and should only contain files which are pertinent to Ziff-Davis' NetBench. Files located in the directory will be overwritten if they correspond to filenames reserved by NetBench. 2. Insert Ziff-Davis' NetBench disk into the floppy drive and copy all NetBench files to the newly created file directory. 3. Run the NetBench License Utility. NETBENCH / L You are required to provide your Name and Organization under the terms of the Licensing Agreement. Follow the directions on the display. 4. Create a file named WS.ID with any text editor and type a unique 3 digit numeric identifier (example: 001) at the first position in the file (no lead spaces or line feeds), then save the file to the C:\WS.ID local drive root directory (Refer to: Definitions WS.ID and Test Stations). The purpose of the file is to identify the test station with a unique 3 digit numeric identifier. NetBench requires this file be located on the C:\ drive root directory of each test station. 5. Once NetBench is installed it is best to: A) Determine which NetBench test to run (NIC Throughput or I/O Throughput) B) Read the corresponding section in this guide about the test. C) If you were actually using the network for testing, you would normally make specific adjustments to the test network at this point in time. Because you have installed NetBench on your local drive, NetBench results are "simulated" and no network adjustments are required. D) Go to the Running NetBench section to learn how to use the menus and configure NetBench parameters for the test. It would be useful to "play" with the test parameters in order to become familiar with them. E) Configure the NetBench parameters F) Run the test. Installing NetBench on a Network Server WARNING The user is cautioned against running these tests on a "live" production network. NetBench is designed to exercise the network in various ways. Running NetBench should be done in isolation when other user activity on the network is non-existent. Load all the NetBench diskette files onto the server such that the program can run off a common logical drive with the other test stations on the network. To do this : 1. Log into the server from your workstation over your LAN connection. 2. Create a default directory on the server. This directory will be accessed by all the test stations as a common workspace from which they will read and write. Any directory name will work. In this example we will use NETBENCH. Warning: This directory must be empty and should only contain files which are pertinent to Ziff-Davis' NetBench. Other files may be overwritten if they are contained in this directory. J: {substitute your Network drive name for J in all cases} MD J:\NETBENCH CD J:\NETBENCH 3. Install the NetBench programs. Insert Ziff-Davis' NetBench disk into the floppy drive and copy the floppy files to the server directory you have designated. 4. Run the NetBench License Utility. CD \NETBENCH NETBENCH / L You are required to provide your Name and Organization under the terms of the Licensing Agreement. Follow the directions on the display. 5. At each Test station: Create a file named WS.ID with any text editor and type a unique 3 digit numeric identifier (example: 001) at the first position in the file (no lead spaces or line feeds), then save the file to the C:\WS.ID local drive root directory (Refer to: Definitions WS.ID and Test Stations). The purpose of the file is to identify each test station in the network with a unique 3 digit numeric identifier. A sequential scheme is suggested for convenience (example: 001, 002, 003). NetBench requires this file be located on the C:\ drive root directory of each test station. 6. Set the login sequence on each test station to execute NetBench on startup. Modify AUTOEXEC.BAT and CONFIG.SYS so that the test stations will boot and automatically start up NetBench for the current test run. (See Novell example Figure 2A and 2B.) Automatic boot up is not required, but suggested for convenience. It also prevents other software interfering with the performance tests. 7. Prevent any non-test stations from interfering with the network. Warning: Do not run these tests from the same network that other users are accessing since it will interfere with both the NetBench testing and the user. Any non-test stations may remain physically connected to the network but must be logged off the network server. 8. Ensure that the file server and network LAN is properly configured for the test. This includes: * LAN server network interface card. * All connectivity and cabling. * Software drivers on LAN server and test stations. * Operating system parameters which might affect the test (Disk caching, memory, maximum packet size, minimum packet size). NetBench is now installed. What's Next? Once NetBench is installed it is best to: * Determine which NetBench test you want to run: NIC Throughput or I/O Throughput. * Read the corresponding section in this guide about the test. * Make any test specific adjustments to the test network. * Go to the Running NetBench section to learn how to use the menus and configure NetBench parameters for the test. * Configure the NetBench parameters. * Run the test. Example 2A (below). Example of Test Station files which use a Novell operating system server (Note: The following is only an example. The user must configure the test stations to match the LAN server used.) An example of the contents of AUTOEXEC.BAT : @ECHO OFF PROMPT $p$g PATH C:\DOS;C:\; set TEMP=C:\DOS ipx Novell net5 Novell e: logon sequence to file server login supervisor logon username to file server g: common logical drive cd \NetBench\ common logical directory NetBench start up NetBench An example of the contents of CONFIG.SYS: DEVICE=C:\DOS\SETVER.EXE DEVICE=C:\DOS\HIMEM.SYS DOS=HIGH FILES=30 lastdrive=d An example of WS.ID contained in the root directory of C:\ 12345678901 column position in file 001 GROUPA1 file contents NOTE: The 3 digit number of the WS.ID field must always be unique on each test station. Figure 2B (below). Example of matching Novell server configuration files STARTUP.NCF set minimum packet receive buffers=200 set maximum physical receive packet size=4096 AUTOEXEC.NCF set enable disk read after write verify=OFF set maximum packet receive buffers=2000 set maximum concurrent disk cache writes=100 set immediate purge of deleted files=ON set new service process wait time=0.3 set maximum service processes=40 Chapter 3 Running NetBench Now that NetBench is installed and the licensing procedure has been executed, it is time to begin interacting with NetBench. To run NetBench, enter: NETBENCH If NetBench does not start up: 1. Check to see if the default directory for NetBench is present. 2. Verify that NetBench has been correctly installed. The ZDLPARMS.CFG file is essential for NetBench to operate. See the ZDLPARMS.CFG Appendix. NetBench Displays NetBench will always start up in the following sequence: Display 1 ZIFF-DAVIS Logo Pause Display 2 Licensing display with Copyright and your ID Pause Display 3 "Waiting" display The "Waiting" display is the default display which will appear on the test station after the NetBench command is issued. All the test stations in the waiting mode are waiting for the coordinator which will trigger testing within its current directory path. One test station must be designated as the Test Station Coordinator. The test station coordinator becomes an interactive test station and no longer has the message "Waiting." Any one test station can become the test coordinator station. The Test Station Coordinator also acts as a test station for the duration of the test and reverts to the Test Station Coordinator again when the test duration has elapsed. Ziff-Davis' NetBench Screensaver This is an optional screen saver "time out" feature which is activated by the Options, General submenu. It "floats" around on the display to prevent "burn- in" of the display phosphor. The default display will reappear by pressing any key. Waiting Display ZD Labs NetBench(TM) . File Performance Options Help Waiting Waiting for test identification from Coordinator station [ OK ] ID:ddd Test No Test Waiting... Group:ooooooo Station Selected ========== =========== Interactive Idle (Words indicated below the === bar indicate status changes which are reflected in the window immediately above.) Activating the Test Station Coordinator Activating the OK box in the waiting window causes the test station to become "interactive" and available to become the Test Station Coordinator. The OK box is selected by pressing the space bar or enter key, or by clicking the OK button with the mouse. Because NetBench is performing activity on the local PC in waiting mode, it is conceivable for the test station to "miss" a mouse click. Repeating the mouse click or using the enter key or space bar will usually assure success. Once the OK button has been activated the "Waiting" window disappears. The word "Interactive" appears in place of "Test Station". It is now possible to select from a menu. To revert back to waiting mode, select the Performance Menu and choose "Wait Mode". While two or more test stations can be in interactive mode, only one test station acting as the Test Station Coordinator may update items within any submenus, or run a test under the Performance Menu. Because there is no practical reason to have multiple test stations in "interactive" mode, Ziff-Davis strongly advises against doing so. Test stations in "interactive" mode cannot be selected as actual test stations except the one actually running the test as the Test Station Coordinator. If a second interactive test station attempts to activate another test concurrently, an error will occur. Basic Navigation - Your choice: Mouse or Keypress Navigation is accomplished by using a mouse. If a mouse is not available or not configured properly, alternative keys may be used. Use the ALT key combined with the LETTER underlined within the menu or list. If a menu is accidentally selected, the ESC key is pressed to deselect the menu. Items may be entered in dialog boxes by typing the requested information. The TAB or SHIFT+TAB or ARROW keys are used to navigate within dialog boxes. The ENTER key commits the action unless otherwise specified. Navigation Keys Mouse - Point and click select a menu Mouse - Point and click activate a choice alt + letter select a menu or choice Esc deselect a sub-menu Tab forward to field Shift+Tab backward to field. Enter commit choice Arrow keys Left, Right highlight a menu item Arrow keys Up, Down highlight a submenu item NetBench Menus and sub-menus File Performance Options Help ____________ ____________ _____________ ______ Data File... NIC Test... Set Name... ReadMe _____________ Batch File... IO Test... General Test... About Stats File... NIC Options... _____________ _____________ IO Options... Batch Run... Wait Mode _____________ _____________ Statistics Stop Loads _____________ Quit Status Message Waiting for test identification from Coordinator station Mode Test name or Error or operation status ID:ddd Group:ooooooo Test No test Waiting.... Station selected ========== ============== ============ Interactive NIC Throughput Polling Test Batch Cmd X I/O Throughput Running Test Test Setup Making Test File Done Aborted Idle (Words indicated below the === bar indicate status changes which are reflected in the window immediately above.) File Menu The file menu allows changes to file names referenced by NetBench. It is also used to invoke the batch run file, and to compute and display statistics captured in the statistics file. Finally, it allows a test station in an interactive mode (test coordinator) to exit the NetBench program back to the DOS operating system. If Data file, Batch File, or Stats File is selected, the user is asked to supply the filename(s) used while running NetBench. Each of the filename parameters has a 12 character limit. The Data file is the data collector file name referenced by NetBench. The data file contains a detailed entry by station identification number (WS.ID) for each test station successfully answering the poll at the conclusion of the test. Each detail line represents the actual "raw" data collected from the test station. It also retains a record of the parameters used during the test. The Data file successively appends each test record to the previous test record. Therefore, consider purging these files periodically. At Ziff-Davis, file naming conventions are used to group tests together. The Stats File (Statistics File) contains a summary of the test. It contains a subset of the configuration data from the Data file. The Stats file also contains a total throughput number that is the sum total of all the test stations reporting at the end of a test run. Every time the Statistics command is issued, the Stats file is regenerated from the current Data file and the total is displayed. The Batch file is used to automate a series of NetBench tests. The batch file will generally take advantage of the group identifier code in the WS.ID file. This section should be skipped until the user becomes familiar with using NetBench. Rules for Batch files The Batch file consists of any number of test statements each terminated by a semi-colon. The file is terminated by a statement consisting of a single exclamation mark. Each statement should begin on a new line. A statement consists of the following items which can be preceded and followed by one or more spaces, tabs or new line characters: Valid Syntax 1) 3 letter test code: NIC or IOT (these must be upper case). 2) Parameter file name (any valid DOS filename except ZDLPARMS.CFG, case insensitive). This parameter file must follow the rules found in Appendix: ZDLPARMS.CFG. 3) One or more group IDs, each of which are no more than 7 characters in length. Group Identifiers correspond to those specified in each individual test station WS.ID file. This field is converted into a DOS file name and is therefore case insensitive. Valid Batch Examples Example: Single Batch Command Command Activity NIC parms.cfg group1; 1. Perform the NIC test, using PARMS.CFG file, activated for those workstations with common group identification of GROUP1 in their WS.ID file ! End of batch file indicator Example: Multiple Batch Commands in a Single File Command Activity NIC parms.cfg group1; 1. Perform the NIC test 2. using PARMS.CFG file 3. activated for those workstations with common group identification of GROUP1 in their WS.ID file IOT test1.prm sector5 sector6 1. Perform the IOT test sector9 ; 2. using a different parameter file called TEST1.PRM 3. using group id's sector 5, sector 6 and sector 9. ! End of batch file indicator Example: Occurrence of Legal Leading and Trailing Spaces Batch File Command Activity ^NIC Perform the NIC test ^whatever.xxx using the file WHATEVER.XXX for parameters ^123 Activated for group 123 ^anygrp^grp2 group anygrp ^onemore group grp2, group onemore ^; end of NetBench statement ^! end of batch file indicator. Note: ^ = one or more spaces The Statistics command must be executed in order to produce the statistics file. Please refer to the comments above (Stats File). The Quit command is executed in order to properly exit NetBench. This command is generally used from the Test Coordinator Station. Stop Loads is a related "exiting" command used from the Test Coordinator Station and is found on the Performance menu. The Quit command exits the test station on which the command is activated while the Stop Loads command exits all test stations that are in the "Waiting" mode. For details please refer to the Stop Loads command. Performance Menu The performance menu contains commands that begin NetBench tests and commands that perform administrative functions. * Selecting NIC Test causes the network interface card test to begin. * Selecting I/O Test causes the I/O throughput test to begin. The parameters associated with each test are supplied from the Options Menu. All test stations perform the same test and each test station displays its individual results on its own monitor. The Test Station Coordinator gathers the data, and saves all results in the Data file. Wait Mode changes a test station from interactive to "waiting". Waiting mode signifies that the test station waits for commands directed to it from the Test Station Coordinator. The command is useful when a test station has accidentally been placed into "interactive" mode. There can only be one Test Station Coordinator for any test on the same network. Stop Loads forces all the test stations to exit the NetBench application and return to the DOS prompt. This option is normally selected when all testing is completed. Use of this command will require the person performing the tests to invoke NetBench on each test station again to resume testing. Once the Stop Loads choice is selected there is no automatic way to resume testing. To continue testing after performing this option requires either rebooting (if the autoexec.bat automatically starts up NetBench), or manually executing NetBench (which starts up in waiting mode). Options Menu The Options menu allows the user to set certain parameters, such as file size and test duration, that influence the test. When a test is completed, these variable parameters are written out to the Data file. The primary categories for this menu are Set Name, General, NIC Options, and I/O. Set Name Set Name allows the user to enter a User Name and a LAN Name. These are treated as comments in the Data file and Stat file (each has a 25 character length limit). General General Test Options allow the user to set Test Length, Variance, Blanking Interval, and Common Path. Test Length (1 - 9999 minutes) Test length, in units of minutes, is the actual duration of the test. This number determines the duration of the test, after any preliminary initialization sequence. Initialization time is not counted as part of this number, nor is polling time counted as part of the test time. Caution: Network tests should be run sufficiently long to reach an ordered and productive behavior pattern (particularly if a large number of work stations are present). If the test is run for an insufficient time, the results may be meaningless. Ziff-Davis usually sets this parameter at 10. Variance (1 - 99 seconds) Variance is a user-specified estimate of time (in seconds) between the first and the last work station completing the test and reporting results. The more test stations that are on the network performing the test, the greater the variance time needed to insure collection of results. Ziff-Davis recommends that this number be set to a minimum of 20 seconds, because most systems will have no more than a 20 second variance when writing results files. If the number of test stations reported in the Data file or Stat file is less than expected, increase the variance number. It may also be necessary to increase this number if the following message is received: "Warning: Not all Load Stations statistics gathered". Blanking Interval (1 - 99 minutes) Blanking interval sets the duration, in minutes, before the screensaver will activate. When the screensaver activates, the display will blank and the Ziff- Davis logo will float around the display. Any key aborts the screensaver and resumes the default NetBench display. Common Path The Common Path parameter specifies the directory that NetBench uses to create temporary files and other temporary test resources. Ziff-Davis recommends that this parameter always be set to ".". NIC Options NIC (Network Interface Card) Options relate specifically to the NIC throughput test. The only available option for this test is the NIC block size. The purpose of this option is to permit a parameter setting of "usable" characters that can be sent through the communication path. NIC Block Size (1 - 65,532 bytes) This number is the size of the blocks in bytes that will be transferred during the NIC Throughput test. Because all test stations are reading from the same file (NETBENCH.EXE), the entire test will run from the cache server subsystem. The test stations will not access the disk and the bottleneck is moved to the NICs. Note: If a network monitor were used to isolate the message sizes sent, the actual packet size would be higher due to the overhead required for the packet routing. The tester needs to be aware of the physical minimum and maximum sizes of usable bytes for the network protocol chosen. Rules for selecting the proper number: For maximum line utilization, this number should be set to the maximum packet-frame size (subtracting the encapsulating header and trailer portion) which your LAN can handle. (Example: 1460 to 1500 for Ethernet; Ethernet maximum including overhead is 1512) For the maximum frames per second, set this number to 1. Note: In order to run this test correctly the parameter entered here must be less than the cache size of the server. I/O I/O Options relate to the Input/Output Throughput test. They exist to help approximate network activity according to a desired model. Four buttons and three parameters permit creation of the desired I/O operation and seek type (random or sequential), I/O file size, I/O block size, and a ratio of reads to writes. Test type has two buttons: Random or Sequential Read / Write has two buttons: Read only, Do Writes. There are three numeric parameters allowable: R/W Ratio, I/O File Size, and I/O Block Size. Valid combinations for these buttons and parameters are represented in the following table: Valid I/O Read only Do Writes Throughput test button button options Sequential Sequential reads Sequential writes button only only R/W Ratio is not R/W Ratio is not used used Random Random reads only Random reads and button writes according R/W Ratio is not to the R/W Ratio used R/W Ratio Parameter Parameter parameter not used enterable and active with Random and Do Writes 0 = write only 1 = 1 read for 1 write 10 = 10 reads to 1 write I/O File Size always required always required parameter I/O Block Size always required always required parameter R/W Ratio (0 - 99 valid range) The Read/Write ratio number determines the number of reads for every write performed during the I/O Throughput test. Setting this at 0 performs writes only. Example: a ratio of 10 indicates that for each write performed on the server there will be 10 reads. Correspondingly a ratio of 1 indicates that for every write performed there will be 1 read. The parameter is only valid for the button combination of Random, and Do Writes. It is ignored for the other tests. I/O File Size (Kb) (1 - 2,097,152 K bytes) The I/O File Size field contains the size of a test file created by each test station during the initialization sequence of the I/O Throughput test. This file is not reused; it is recreated each time the test is run. A larger number will increase the disk surface area on which the test operates, placing the server disk subsystem under more stress. Allocating a large file size is desirable as it reduces the impact of caching on test performance. If a large number of test stations are present and disk space on the file server is low, the file size may need to be reduced if the test file creation fails. I/O Block Size (Bytes) (1 - 65532 bytes) The I/O block size field contains the size (in bytes) of the block to be read or written by the test station during the I/O Throughput test. This block should be significantly smaller than the test file to allow for random I/O or due to caching, the results may be meaningless. Help The HELP option permits access to a ReadMe ASCII file and an About statement. The ReadMe file contains the Menu and Sub-menu portion of the NetBench User Guide. The About statement contains the product name, version, program developers, and the Ziff-Davis staff at the time of release. Chapter 4 Network Interface Card (NIC) Throughput Test This chapter describes how to use the Network Interface Card (NIC) Throughput Test. Mechanics During this test, all NetBench test stations perform sequential reads from the NETBENCH.EXE file of the size specified by NIC Throughput Block Size. Because all stations are reading from the same file, the entire test will run from the server cache subsystem. Because NETBENCH.EXE is in the server cache, the test stations will not access the disk and the maximum data throughput will be limited by the throughput capability of the network interface cards. The test stations count the number of blocks read from the server during the test and display the total throughput (number of iterations times the block size in bytes) when the test is over. Objective The objective of this test is to exercise network activity from the NIC test station through the NIC on the file server. This test result indicates the maximum amount of network traffic which can occur within a specific amount of time using the test stations supplied at a specific block size. (Refer to the Running NetBench section, Parameters, Testing the server NIC and Testing the test station NIC). The NIC Throughput test results are in bits per second Factors that influence the test * NIC card contained in the server and associated driver software. * NIC card contained in the test station and associated driver software. * Cache on NIC card. * Theoretical Maximum network speed (Ethernet 10 Mbits, Token Ring 4 Mbits, 16 Mbits, FDDI, etc.). * Cable type - bandwidth. * CPU in the server. * CPU in the test station. * Memory in server. * Memory in test station. Parameters NIC Throughput Block Size. - The user should vary the block size to simulate the network application for small and large block sizes for successive test runs. Workstation drivers usually have a minimum block size. The parameter selected for NetBench should never drop below this minimum because network activity must be forced to occur on the cable used. Ridiculous results may be an indicator of "out of bounds" conditions, for example: 15 Mbits / sec when the maximum for Ethernet is 10 Mbits /sec. Special Rules 1. This test is network operating system (NOS) dependent, in that one NOS cannot properly be compared with another NOS and also vary the NIC at the same time. 2. Do not turn cache memory off on the server. 3. Network configuration files in the test station must remain constant for a proper comparison of NIC cards. 4. Different software driver versions from the same manufacturer may affect the results. Be certain the item being measured is the correct one. Vendor cards may also have version differences. Guidelines Determine which type of card to test. There are often two kinds of NIC adapters: server cards and workstation cards. The NIC installed in the server must have as high or higher performance capability than the test station NIC for the test to be informative. Testing the server NIC If the test is designed to test the server network interface card throughput, replace only that card in the server and rerun the test with each new server card to be tested, not varying the NIC adapters in the test stations. This type of test would generally require a server and multiple test stations. Example: For testing an Ethernet Novell network, 1 server and 4 test stations are probably required. If each test station had the capacity to generate 3 Mbits of information the network would be saturated, because 4 stations times 3 Mbits = 12 Mbits and the capacity of the Ethernet is only 10 Mbits. Testing the test station NIC If your test is designed to evaluate the test station NIC, only one server and one test station is required. The highest performance NIC card would still be placed in the server for the test. However the various cards that were being evaluated would be placed in the test station, varying one at a time for each test with all other factors remaining the same. Setting up and Running the Test 1. Configure the physical network, cable connectivity and server connectivity. Ensure that the network is in operation, proper software network parameters have been established, NetBench has been installed on the server common drive from a workstation and licensed, and that all test stations have been logged on the server and have changed their current directory to the common drive on the server. Each test station should start up NetBench using this common drive and directory. The test stations must not reference NetBench from any other location (such as a copy on a local disk). 2. Select one test station to act as the Test Station Coordinator. 3. Set the File Menu parameters (File Menu - Data File, Stats File). 4. Set the Options (Options - Set Name, General Test, NIC options) 5. Run the test (Performance, NIC test) 6. The display on each test station should show its individual results. This will also be true of the Test Station Coordinator. 7. Collect the results (File menu - Statistics). A summary will display on the Test Station Coordinator. The total number of test stations responding to the test will be shown in the Final Statistics window. 8. Exit NetBench (File - Quit). (Or run the next test.) 9. Examine the detail results in the Data File using a text editor or from a printout. There should be an entry in this file for each test station which reported back results to the Test Station Coordinator during the polling sequence. The entry is listed by the station WS.ID number. If an entry is missing, there are three possible causes. Cause A The missing test station was not setup correctly pointing to the common directory on the file server. Cause B The WS.ID file is incorrect or has the same number on two or more of the test stations. These must be unique numbers for each test station on the network and must be on the individual C: root directory for each test station. Cause C The Variance setting under Options Menu, General Test was too short for all the test stations to report back in sufficient time to the Test Station Coordinator. Increase this value. However, if there are fewer than 10 test stations on the network and the variance setting is greater than 30 seconds, re-examine Cause A or B, and the network connectivity. Thirty seconds should be plenty of time for 10 stations to respond unless something is wrong with the network itself. 10. Examine the totals for all test stations in the Statistics File. Evaluation Scenarios Example 1 The test bed consisted of one test station connected to a Novell Netware server. Vendor XYZ 486/33M systems with 12MB of RAM was used for the server and Vendor XYZ 386/25 with 8MB of RAM was used for the test station. The server was loaded with Novell's Netware 3.11 network operating system. The test station was a DOS client loaded with ABC drivers. A QRX card was installed in the server and remained in place throughout this test. Each card tested was placed in the workstation in turn. Using Ziff-Davis' NetBench NIC test, the performance of the card was measured at different block sizes (512, 1K, 2K bytes). The test was performed for the VendorCardA, VendorCardB, VendorCardC, and VendorCardD cards. The duration of each test was 10 minutes. Test 1 (Netware Server) Vendor Name 512B 1KB 2KB VendorCard A 2331989 3402410 3899939 VendorCard B 1889621 2747050 3064162 VendorCard C 2009770 2861738 3200341 VendorCard D 2012501 2824263 3145728 Example 2 The test bed consisted of 3 test stations and one LANManager server. Ziff-Davis used Computer 486/33M systems with 12MB of RAM for the server and 486/33M with 8MB of RAM for the workstations. The server was loaded with Microsoft's LANManager 2.1 with OS/2 1.3 as a primary operating system. The test stations used DOS. A vendor ACF card was installed in each of the three client test stations and in the server. Using Ziff-Davis’ NetBench NIC test, the total throughput was measured at a 4Kbyte block size. The same procedure was repeated by replacing the ACF card in the server with the FGH card. The duration of each test was 10 minutes. VendorName Throughput(bits per sec) FGH 9209874 ACF (server) 8312173 ACF (client) 8194716 Chapter 5 I/O Throughput Test This chapter explains how to use the I/O Throughput Test. Mechanics When this test is selected, each test station writes a test file to the common server drive. Once the Test Station Coordinator determines that all test stations have completed this initialization operation, I/O throughput testing begins. During the test, each test station performs the file reads and/or writes (at the block size, for the length of time specified, according to the parameters set under the I/O Options menu). After the test duration has elapsed, each test station reports its results. Objective The objective of this test is to determine the performance of a file server relative to a baseline system. Performance in this case is defined in terms of two concepts - peak I/O throughput, and scalability. Peak I/O throughput is the maximum amount of user data (block size and file size) which can be read or written to the server during the test. Scalability is the ability of a server to sustain peak I/O throughput (disk and disk cache) as the file service load on the server is increased during the test. File server performance should be measured across the test station to the server disk path. To use this test effectively, ensure that the system is configured so that the desired components are stressed and that other extraneous components are eliminated with respect to their influence in such a way that they do not hamper performance. Components, such as the number of clients and the network wire must be set up in such a way as not to cause bottlenecks. For example, to prevent the NIC from becoming the bottleneck, multiple NICs would be placed in the server, each corresponding to a separate LAN segment (usually four NICs are sufficient). Multiple test stations and a server NIC would be attached to each segment so that the server becomes a common link to all LAN segments. Additional test stations could be added to the segments and the test re-run to identify the point at which server I/O Throughput begins to be maximized. Also, by using higher bandwidth network services, such as 16Mbit Token Ring, the network bottleneck is reduced and places more emphasis on file service performance. This will cause the disk subsystem to be stressed rather than the LAN NICs. To use NetBench properly, a throughput curve should be plotted for comparing performance measurement. A throughput curve is formed by combining the results (datapoints) from multiple NetBench iterations into a single dataset. The dataset is formed by configuring a baseline network, setting the I/O Options parameters, running the I/O throughput test, varying one load factor (for example, increasing test stations) and running the test again, varying the same load factor and running the test again, etc. A throughput curve, consisting of charting the collection of data points, not only displays the peak throughput of a server, but also demonstrates the servers scalability and how the server will perform from light to harsh conditions. By comparing one single data point to multiple points (the throughput curve), it becomes clear that a single point measurement can be very misleading. For example, experiments with this test have shown that throughput curves of servers being compared often cross. A server which displays superior performance at low test station loads may display inferior performance as the load is increased. This information is very important to a complete analysis of performance. Throughput curves for file servers generally fall into three categories. The most common is the Poor Scalability category, typical of a desktop machine turned into a file server. This curve may display an acceptable throughput at a light test station load, but the performance quickly drops off as the load is increased. The next category is that of Median Scalabilty. This curve has a very high peak throughput but as soon as the client load rises above the ability of the cache, the performance drops rapidly and bottoms out at a low level. The last is the Superior Scalability category. This type of curve has a very high peak throughput that is sustained far beyond the other two categories. I/O Throughput test results are in bits/ second. Factors that influence the test * Cache in the server. * Cache in the test station. * Number of test stations. * All NetBench Option parameters for I/O Options. Parameters I/O File Size (in Kbytes) One disk file is created for each test station prior to the commencement of the test (of the size specified in I/O Throughput File Size). This file is created on the server in the current working directory. Increasing this number creates a greater amount of relative disk head movement from the beginning of the file to the end of the file. I/O Block Size (in Bytes) Each test station testfile is read or written a block at a time according to the I/O Throughput Block Size parameter. Increasing this number decreases the amount of disk activity for the given test (if other factors are held constant). Read or Write and Random vs Sequential are specified according to radio buttons selected in the Options, I/O Options submenu. There are five read/write options for this test. These are Sequential Read only, Sequential Write only, Random Read only, Random Write only*, and Random Read/Write*. If Sequential is selected, the only testing performed is Sequential Read or Sequential Write according to the respective button selected: Read Only or Do Writes. Sequential Read and Write is NOT performed. If Random is selected with the Read Only button the only testing performed are Random Reads. If Random is selected with the Do Writes button, then Random Reads and Random Writes are permitted according to the number specified in R/W Ratio field. I/O Options Definitions Sequential means that each test file created on the server by the test station is read or written one block at a time according to the Read or Write operation selected. When the end of the file is reached, the test sequence starts over and continues until the test duration limit has elapsed. Random means that each test file created on the server by each test station is read or written one block at a time according to the Read, Write, Read/Write operation selected. The relative "start point" of the read or write is randomly determined by each test station. This process repeats until the test duration has elapsed. An attempt to read or write beyond the end of the file size allocated is not permitted. * When both the Random button and Do Writes button are selected, the mix of read/write operations are determined by the read/write ratio field. The read/write ratio field represents the number of read operations for every write operation. Example: If the ratio is 3, there are 3 reads for every 1 write. If the ratio is 0 there is only random writing. If the ratio is 10, there are 10 reads for 1 write. Guideline Generally use multiple NICs in the server with multiple LAN segments in order to cause a bottleneck at the disk. The number of LAN segments may also be varied. Setting up and Running the Test 1. Configure the physical network, cable connectivity and server connectivity. Ensure that the network is in operation, proper software network parameters have been established, NetBench has been installed on the server common drive from a workstation and licensed, and that all test stations have been logged on the server and have changed their current directory to the common drive on the server. Each test station should start up NetBench using this common drive and directory. The test stations must not reference NetBench from any other location (such as a copy on a local disk). 2. Select one test station to act as the Test Station Coordinator. 3. Set the File Menu parameters (File Menu - Data File, Stats File). 4. Set the Options (Options - Set Name, General Test, IO options). 5. Run the test (Performance, IO test). 6. The display on each test station should show its individual results. This will also be true of the Test Station Coordinator. 7. Collect the results (File menu - Statistics). A summary will display on the Test Station Coordinator. The total number of test stations responding to the test will be shown in the Final Statistics window. 8. Exit NetBench (File - Quit). (Or run the next test.). 9. Examine the detail results in the Data File using a text editor or from a printout. There should be an entry in this file for each test station which reported back results to the Test Station Coordinator during the polling sequence. The entry is listed by the station WS.ID number. If an entry is missing, there are three possible causes. Cause A The missing test station was not setup correctly pointing to the common directory on the file server. Cause B The WS.ID file is incorrect or has the same number on two or more of the test stations. These must be unique numbers for each test station on the network and must be on the individual C: root directory for each test station. Cause C The Variance setting under Options Menu, General Test was too short for all the test stations to report back in sufficient time to the Test Station Coordinator. Increase this value. However, if there are fewer than 10 test stations on the network and the variance setting is greater than 30 seconds, re-examine Cause A or B, and the network connectivity. Thirty seconds should be plenty of time for 10 stations to respond unless something is wrong with the network itself. 10. Examine the totals for all test stations in the Statistics File. Evaluation Scenarios Example of evaluating servers at 1024 K bytes I/O throughput evenly distributed on 4 NIC cards. Two NICs are Token Ring, Two use Ethernet. # of Test Stations Server Vendor C Server Vendor D Server Vendor E 8 17317884 12937894 14231139 16 23567830 18288905 14905063 24 24918961 20149030 15439179 48 23426979 19689402 12576168 72 21369669 16986887 4200039 96 18114029 13708320 3880144 Example of evaluating servers with 96 test stations evenly distributed on 4 NICs. Two NICs are 16 Mbit Token Ring, Two NICs are Ethernet. File size Kbytes Server Vendor C Server Vendor D Server Vendor E 512 23464845 19857675 6469524 1024 18114029 13708320 3880144 2048 14063785 10256700 2980629 4096 12039712 9155468 2635356 8192 10873353 7716024 2319071 These statistics were gathered using Novell 3.11 using the same amount of RAM. Different CPU, and disks were used. Test Length was 10 minutes. Appendix A Data File and Statistics File Parameters Parameters Common to Both Files * @@@ * Date / Time Stamp. * User name comment line. * LAN Tested comment line Test File Options, I/O (I/O throughput Options, I/O File test) kbytes Size IO Blocks Options, I/O (I/O throughput Options, I/O test) bytes Blocks Do Writes Options, I/O (I/O throughput Options test) Y=Do Writes, N=Read Only Sequential Options, I/O (I/O throughput Options test) Y=Sequential N=Random R/W Ratio Options, I/O (I/O throughput Options test) NT Blocks Options, NIC (NIC Throughput Options, NIC Test) bytes Block Size Test Length Options, General Options Variance Options, General Options Blank Int. Options, General Blanking Interval Options Common Path Options, General Options Data File File, Data Batch File File, Batch Comment Performance, (any test selected) Test Type Performance, (any test selected) NIC Throughput ... { Common parameters }... @@@ 012 3283690 0 123200 600 022 3283690 0 123200 600 132 3283690 0 123200 600 @@@ @@@ Indicates start or end of repeating data. Each line represents results from each test station. The test station fields are described from left to right. * Workstation ID * Test Station Throughput (bits/sec) * Total Writes During the Test (Always 0) * Total Reads During the Test * Duration of the Test Station run (sec) I/O Throughput ... { Common parameters } ... @@@ 010 143360 0 1050 60 0.00000 0.05714 011 143360 0 1050 60 0.00000 0.05714 012 143360 0 1050 60 0.00000 0.05714 @@@ @@@ Indicates start or end of repeating data. Each line represents results from each test station. The test station fields are described from left to right. * Workstation ID * Test Station I/O Throughput (bits/sec) * Total Writes During the Test * Total Reads During the Test * Duration of the Test Station run (sec) * Average Time Per Write Cycle(sec)* * Average Time Per Read Cycle(sec)* *Note that the value for these two items will always be 0.00000 when the Random and Do Writes I/O Throughput test options are selected, and the Read Write Ratio is greater than 0. Stats File This file contains the configuration information from each run accompanied by an extra line that has the total system throughput for that test iteration. Appendix B Temporary Files NetBench creates a number of temporary files for coordination between test stations. The names ( or "wild card" syntax) are listed below. This is not a comprehensive list. The user should avoid placing any file in this directory that is important to retain. IOTEST .SEM NTTEST .SEM CBTEST .SEM NOTEST .SEM ??TEST .SEM EXTEST .SEM IOT_YES .SEM IOT_NO .SEM IOT_* .SEM FILEOPT .SEM EVALOPT .SEM STATOPT .SEM STS* .SEM STF* .SEM STB* .SEM ST* .SEM B* .SEM YXZVQ???.X$X IOT* .X$X * .X$X Appendix C ZDLPARMS.CFG Listing of ZDLPARMS.CFG (same format for any Parameter file) Mike Bicuspid User Name Lan A LAN Name 1024 I/O File Size Kbytes 512 I/O Block Size Bytes 1 1=Do Writes; 0=Read only 0 1=Sequential; 0=Random 3 R/W Ratio 1999 NIC Block size bytes 10 Test Length minutes 20 Variance in seconds 5 Blanking Interval minutes . Current Directory (period) data.dat Data File batch.txt Batch File stat.dat Stats File NetBench(tm) Version 2.10 The ZDLPARMS.CFG file is essential for NetBench to operate. If this file is accidentally deleted, NetBench cannot run. If the ZDLPARMS.CFG file is missing, obtain it from the backup disk, or look for an alternative file PARMS.CFG. Copy the ".cfg" file to ZDLPARMS.CFG. NetBench will then startup properly. If this file cannot be located, the file can be carefully created with an editor. The New line character terminates each line. It is essential that the new line character appears within the length limitation set forth in the documentation. This file is in the same format required for the batch parameter files listed in the batch file syntax. For batch operations, it is quickest to copy the ZDLPARMS.CFG file to a new file name and then carefully edit the parameters you wish to change. Alternatively, set up your parameters within NetBench, then quit and copy ZDLPARMS.CFG to a new file name. Repeat this procedure for each new parameter file. Appendix D Data File Output Example @@@ Sat Sep 26 16:52:14 1992 User Name : Mike Bicuspid LAN Tested : Token Ring Test File : 1024 kb IO Blocks : 8192 b Do Writes : Y Sequential : N R/W Ratio : 3 NT Blocks : 512 b Test Length: 10 minutes Variance : 20 seconds Blank Int. : 15 minutes Common Path: . Data File : nb1912.dat Batch File : batch.txt Stats File : nb1912s.dat Comment : NetBench test Test Type : NIC Throughput @@@ 026 1611244 0 35600 181 0.00000 0.00000 027 380181 0 8400 181 0.00000 0.00000 028 537180 0 12000 183 0.00000 0.00000 029 504123 0 11200 182 0.00000 0.00000 030 452596 0 10000 181 0.00000 0.00000 036 1529173 0 33600 180 0.00000 0.00000 037 133565 0 6000 184 0.00000 0.00000 038 543116 0 12000 181 0.00000 0.00000 039 468114 0 10400 182 0.00000 0.00000 040 432105 0 9600 182 0.00000 0.00000 @@@ Sat Sep 26 16:57:21 1992 User Name : Mike Bicuspid LAN Tested : Token Ring Test File : 1024 kb IO Blocks : 8192 b Do Writes : Y Sequential : N R/W Ratio : 3 NT Blocks : 512 b Test Length: 10 minutes Variance : 20 seconds Blank Int. : 15 minutes Common Path: . Data File : nb1912.dat Batch File : batch.txt Stats File : nb1912s.dat Comment : NetBench test Test Type : I/O Throughput @@@ 026 1194211 820 2460 180 0.2195 0.00000 027 728177 500 1500 180 0.3600 0.00000 028 833035 572 1716 180 0.3147 0.00000 029 766043 526 1578 180 0.3422 0.00000 030 798082 548 1644 180 0.3285 0.00000 036 1188386 816 2448 180 0.2206 0.00000 037 570633 394 1182 181 0.4594 0.00000 038 734003 504 1512 180 0.3571 0.00000 039 805260 556 1668 181 0.3255 0.00000 040 675748 464 1392 180 0.3879 0.00000 @@@ Sat Sep 26 17:01:41 1992 User Name : Mike Bicuspid LAN Tested : Token Ring Test File : 1024 kb IO Blocks : 8192 b Do Writes : Y Sequential : N R/W Ratio : 3 NT Blocks : 512 b Test Length: 10 minutes Variance : 20 seconds Blank Int. : 15 minutes Common Path: . Data File : nb1912.dat Batch File : batch.txt Stats File : nb1912s.dat Comment : NetBench test Test Type : CPU Bandwidth @@@ 026 171 30901 180 0.00000 0.00000 027 80 14401 180 0.00000 0.00000 028 73 13201 180 0.00000 0.00000 029 81 14601 180 0.00000 0.00000 030 84 15201 180 0.00000 0.00000 036 136 24701 181 0.00000 0.00000 037 80 14401 180 0.00000 0.00000 038 96 17401 180 0.00000 0.00000 039 90 16501 182 0.00000 0.00000 040 87 15801 181 0.00000 0.00000 @@@ Sat Sep 26 17:06:06 1992 User Name : Mike Bicuspid LAN Tested : Token Ring Test File : 512 kb IO Blocks : 8192 b Do Writes : Y Sequential : N R/W Ratio : 3 NT Blocks : 1024 b Test Length: 10 minutes Variance : 20 seconds Blank Int. : 15 minutes Common Path: . Data File : nb1912.dat Batch File : batch.txt Stats File : nb1912s.dat Comment : NetBench test Test Type : NIC Throughput @@@ 026 1419946 0 31200 180 0.00000 0.00000 027 414101 0 9200 182 0.00000 0.00000 028 525012 0 11600 181 0.00000 0.00000 029 519274 0 11600 183 0.00000 0.00000 030 418702 0 9200 180 0.00000 0.00000 036 1456355 0 32000 180 0.00000 0.00000 037 425098 0 9600 185 0.00000 0.00000 038 582542 0 12800 180 0.00000 0.00000 039 468114 0 10400 182 0.00000 0.00000 040 455111 0 10000 180 0.00000 0.00000 @@@ Sat Sep 26 17:11:11 1992 User Name : Mike Bicuspid LAN Tested : Token Ring Test File : 512 kb IO Blocks : 8192 b Do Writes : Y Sequential : N R/W Ratio : 3 NT Blocks : 1024 b Test Length: 10 minutes Variance : 20 seconds Blank Int. : 15 minutes Common Path: . Data File : nb1912.dat Batch File : batch.txt Stats File : nb1912s.dat Comment : NetBench test Test Type : I/O Throughput @@@ 026 1193406 824 2472 181 0.2197 0.00000 027 728177 500 1500 180 0.3600 0.00000 028 780606 536 1608 180 0.3358 0.00000 029 798082 548 1644 180 0.3285 0.00000 030 675748 464 1392 180 0.3879 0.00000 036 1127219 774 2322 180 0.2326 0.00000 037 710701 488 1464 180 0.3689 0.00000 038 798082 548 1644 180 0.3285 0.00000 039 835948 574 1722 180 0.3136 0.00000 040 771868 530 1590 180 0.3396 0.00000 @@@ Sat Sep 26 17:15:31 1992 User Name : Mike Bicuspid LAN Tested : Token Ring Test File : 1024 kb IO Blocks : 8192 b Do Writes : Y Sequential : N R/W Ratio : 3 NT Blocks : 512 b Test Length: 10 minutes Variance : 20 seconds Blank Int. : 15 minutes Common Path: . Data File : nb1912.dat Batch File : batch.txt Stats File : nb1912s.dat Comment : NetBench test Test Type : NIC Throughput @@@ 026 1601991 0 35200 180 0.00000 0.00000 028 540131 0 12000 182 0.00000 0.00000 029 561219 0 12400 181 0.00000 0.00000 030 527928 0 11600 180 0.00000 0.00000 036 1683659 0 37200 181 0.00000 0.00000 038 582542 0 12800 180 0.00000 0.00000 039 543116 0 12000 181 0.00000 0.00000 040 546133 0 12000 180 0.00000 0.00000 @@@ Sat Sep 26 17:19:51 1992 User Name : Mike Bicuspid LAN Tested : Token Ring Test File : 1024 kb IO Blocks : 8192 b Do Writes : Y Sequential : N R/W Ratio : 3 NT Blocks : 512 b Test Length: 10 minutes Variance : 20 seconds Blank Int. : 15 minutes Common Path: . Data File : nb1912.dat Batch File : batch.txt Stats File : nb1912s.dat Comment : NetBench test Test Type : NIC Throughput @@@ 026 1818443 0 40400 182 0.00000 0.00000 029 720175 0 16000 182 0.00000 0.00000 030 669843 0 14800 181 0.00000 0.00000 036 1792282 0 39600 181 0.00000 0.00000 039 819200 0 18000 180 0.00000 0.00000 040 651739 0 14400 181 0.00000 0.00000 @@@ Sat Sep 26 17:24:11 1992 User Name : Mike Bicuspid LAN Tested : Token Ring Test File : 1024 kb IO Blocks : 8192 b Do Writes : Y Sequential : N R/W Ratio : 3 NT Blocks : 512 b Test Length: 10 minutes Variance : 20 seconds Blank Int. : 15 minutes Common Path: . Data File : nb1912.dat Batch File : batch.txt Stats File : nb1912s.dat Comment : NetBench test Test Type : NIC Throughput @@@ 026 2038897 0 44800 180 0.00000 0.00000 030 1037653 0 22800 180 0.00000 0.00000 036 2166328 0 47600 180 0.00000 0.00000 040 1134276 0 25200 182 0.00000 0.00000 @@@ Sat Sep 26 17:29:11 1992 User Name : Mike Bicuspid LAN Tested : Token Ring Test File : 512 kb IO Blocks : 8192 b Do Writes : Y Sequential : N R/W Ratio : 3 NT Blocks : 1024 b Test Length: 10 minutes Variance : 20 seconds Blank Int. : 15 minutes Common Path: . Data File : nb1912.dat Batch File : batch.txt Stats File : nb1912s.dat Comment : NetBench test Test Type : I/O Throughput @@@ 026 1808793 1242 3726 180 0.1449 0.00000 029 1179648 810 2430 180 0.2222 0.00000 030 1034012 710 2130 180 0.2535 0.00000 036 1721412 1182 3546 180 0.1523 0.00000 039 1034012 710 2130 180 0.2535 0.00000 040 1112655 764 2292 180 0.2356 0.00000 @@@ Sat Sep 26 17:34:11 1992 User Name : Mike Bicuspid LAN Tested : Token Ring Test File : 512 kb IO Blocks : 8192 b Do Writes : Y Sequential : N R/W Ratio : 3 NT Blocks : 1024 b Test Length: 10 minutes Variance : 20 seconds Blank Int. : 15 minutes Common Path: . Data File : nb1912.dat Batch File : batch.txt Stats File : nb1912s.dat Comment : NetBench test Test Type : I/O Throughput @@@ 026 2236962 1536 4608 180 0.1172 0.00000 030 1430141 982 2946 180 0.1833 0.00000 036 2271914 1560 4680 180 0.1154 0.00000 040 1517522 1042 3126 180 0.1727 0.00000 Appendix E Statistics File Output Example Sat Sep 26 16:52:14 1992 User Name : Mike Bicuspid LAN Tested : Token Ring Test File : 1024 kb IO Blocks : 8192 b Do Writes : Y Sequential : N R/W Ratio : 3 NT Blocks : 512 b Test Length: 10 minutes Variance : 20 seconds Blank Int. : 15 minutes Common Path: . Data File : nb1912.dat Batch File : batch.txt Stats File : nb1912s.dat Comment : NetBench test Test Type : NIC Throughput Total Throughput = 6591397 Number of Work Stations = 10 Sat Sep 26 16:57:21 1992 User Name : Mike Bicuspid LAN Tested : Token Ring Test File : 1024 kb IO Blocks : 8192 b Do Writes : Y Sequential : N R/W Ratio : 3 NT Blocks : 512 b Test Length: 10 minutes Variance : 20 seconds Blank Int. : 15 minutes Common Path: . Data File : nb1912.dat Batch File : batch.txt Stats File : nb1912s.dat Comment : NetBench test Test Type : I/O Throughput Total Throughput = 8293578 Number of Work Stations = 10 Sat Sep 26 17:01:41 1992 User Name : Mike Bicuspid LAN Tested : Token Ring Test File : 1024 kb IO Blocks : 8192 b Do Writes : Y Sequential : N R/W Ratio : 3 NT Blocks : 512 b Test Length: 10 minutes Variance : 20 seconds Blank Int. : 15 minutes Common Path: . Data File : nb1912.dat Batch File : batch.txt Stats File : nb1912s.dat Comment : NetBench test Test Type : CPU Bandwidth Total Throughput = 978 Number of Work Stations = 10 Sat Sep 26 17:06:06 1992 User Name : Mike Bicuspid LAN Tested : Token Ring Test File : 512 kb IO Blocks : 8192 b Do Writes : Y Sequential : N R/W Ratio : 3 NT Blocks : 1024 b Test Length: 10 minutes Variance : 20 seconds Blank Int. : 15 minutes Common Path: . Data File : nb1912.dat Batch File : batch.txt Stats File : nb1912s.dat Comment : NetBench test Test Type : NIC Throughput Total Throughput = 6684255 Number of Work Stations = 10 Sat Sep 26 17:11:11 1992 User Name : Mike Bicuspid LAN Tested : Token Ring Test File : 512 kb IO Blocks : 8192 b Do Writes : Y Sequential : N R/W Ratio : 3 NT Blocks : 1024 b Test Length: 10 minutes Variance : 20 seconds Blank Int. : 15 minutes Common Path: . Data File : nb1912.dat Batch File : batch.txt Stats File : nb1912s.dat Comment : NetBench test Test Type : I/O Throughput Total Throughput = 8419837 Number of Work Stations = 10 Sat Sep 26 17:15:31 1992 User Name : Mike Bicuspid LAN Tested : Token Ring Test File : 1024 kb IO Blocks : 8192 b Do Writes : Y Sequential : N R/W Ratio : 3 NT Blocks : 512 b Test Length: 10 minutes Variance : 20 seconds Blank Int. : 15 minutes Common Path: . Data File : nb1912.dat Batch File : batch.txt Stats File : nb1912s.dat Comment : NetBench test Test Type : NIC Throughput Total Throughput = 6586719 Number of Work Stations = 8 Sat Sep 26 17:19:51 1992 User Name : Mike Bicuspid LAN Tested : Token Ring Test File : 1024 kb IO Blocks : 8192 b Do Writes : Y Sequential : N R/W Ratio : 3 NT Blocks : 512 b Test Length: 10 minutes Variance : 20 seconds Blank Int. : 15 minutes Common Path: . Data File : nb1912.dat Batch File : batch.txt Stats File : nb1912s.dat Comment : NetBench test Test Type : NIC Throughput Total Throughput = 6471682 Number of Work Stations = 6 Sat Sep 26 17:24:11 1992 User Name : Mike Bicuspid LAN Tested : Token Ring Test File : 1024 kb IO Blocks : 8192 b Do Writes : Y Sequential : N R/W Ratio : 3 NT Blocks : 512 b Test Length: 10 minutes Variance : 20 seconds Blank Int. : 15 minutes Common Path: . Data File : nb1912.dat Batch File : batch.txt Stats File : nb1912s.dat Comment : NetBench test Test Type : NIC Throughput Total Throughput = 6377154 Number of Work Stations = 4 Sat Sep 26 17:29:11 1992 User Name : Mike Bicuspid LAN Tested : Token Ring Test File : 512 kb IO Blocks : 8192 b Do Writes : Y Sequential : N R/W Ratio : 3 NT Blocks : 1024 b Test Length: 10 minutes Variance : 20 seconds Blank Int. : 15 minutes Common Path: . Data File : nb1912.dat Batch File : batch.txt Stats File : nb1912s.dat Comment : NetBench test Test Type : I/O Throughput Total Throughput = 7890532 Number of Work Stations = 6 Sat Sep 26 17:34:11 1992 User Name : Mike Bicuspid LAN Tested : Token Ring Test File : 512 kb IO Blocks : 8192 b Do Writes : Y Sequential : N R/W Ratio : 3 NT Blocks : 1024 b Test Length: 10 minutes Variance : 20 seconds Blank Int. : 15 minutes Common Path: . Data File : nb1912.dat Batch File : batch.txt Stats File : nb1912s.dat Comment : NetBench test Test Type : I/O Throughput Total Throughput = 7456539 Number of Work Stations = 4 Appendix F Technical Support NetBench is available free of charge on 3 1/2" diskettes. If you would like to receive a copy of NetBench or any other Ziff-Davis benchmark, send your name, address, and phone number to the Ziff-Davis Benchmark Operation (ZDBOp) benchmark request FAX number (919-380-2879) using the form in this appendix. The benchmark will arrive via third-class US Mail. If you want to receive this benchmark sooner and you have a Federal Express account, include your account number and shipping instructions on the FAX form, and the benchmark will arrive via Federal Express. Contacting ZDBOp If you have comments or questions about NetBench, please contact ZDBOp. You can contact ZDBOp in several ways: * If you have a modem and communications software, you can use CompuServe* to access ZiffNet, the Ziff-Davis on-line service. The ZiffNet forum is a way for you to communicate your questions and comments to Ziff-Davis. (Access to CompuServe is available for a fee.) If you are a member of CompuServe, simply type GO ZIFFNET at the CompuServe prompt. If you are not a member of CompuServe, you can access ZiffNet following the steps below: 1. Set your communications software to seven data bits, even parity, one stop bit. Select a data transfer rate (bits per second) of 1200, 2400, or 9600. Using your modem, call the local ZiffNet number. To find your local ZiffNet number, dial 800-346-3247 (by modem) or 800-635-6225 (by voice). If you login by modem, at the user ID prompt, type PHONES. 2. Respond to the prompts by entering the information below: When you connect: Host Name: CIS User ID: 177000,5555 Password: ZIFF*NET Agreement Number: MACMAN2 Register your name and credit card number for billing. 3. You will receive a Personal User ID and Temporary Password on-line. Write them down and use them to log on. You will get a permanent password by mail within 10 business days. * You can FAX your questions and comments directly to NetBench Technical Support at the ZDBOp benchmark FAX number (919-380-2879). * You can mail your questions and comments to ZDBOp at the following address: Ziff-Davis Benchmark Operation One Copley Parkway, Suite 510 Morrisville, North Carolina 27560 Attn.: NetBench Technical Support Benchmark Request Form FAX OR MAIL THIS FORM TO: The Ziff-Davis Benchmark Operation One Copley Parkway, Suite 510 Morrisville, NC 27560 FAX: (919) 380-2879 for assistance call (919) 380-2800 WHO I AM: Name: ______________________________________________________ Company: ____________________________________________________ Address: _____________________________________________________ City: _________________________ State: ________ Zip: ______________ Phone Number: ________________________________________________ Fax: _________________________________________________________ We answer requests in the order that we receive them. We ship all benchmarks via third-class US Mail unless you supply a Federal Express account number. Please allow four to six weeks for delivery via regular mail. Your Federal Express account number: ______________________________ Check one: ___ priority overnight ___standard overnight WHAT I WANT: ___ PC Bench* 8.0 ___ WinBench* 4.0 ___ Winstone* 1.0 (available only on CD-ROM) ___ NetBench* 2.1 ___ ServerBench* 1.0 ___ MacBench* 1.0 (All benchmarks EXCEPT Winstone are available only on 3.5" diskettes.) Glossary Batch Run File This is an optional text file which contains NetBench commands. It is designed to run a series of NetBench tests unattended. NIC Network interface card. Server A computer containing the physical drive that contains the NetBench files. Test stations reference this common logical drive in order to run NetBench. Test station A DOS computer running NetBench. Each computer waits to be triggered by the Test Station Coordinator to begin the test. After the test is completed, the results are collected by the Test Station Coordinator. Each test station is identified with a unique user-assigned identifier called WS.ID. A PC workstation must be placed in NetBench "waiting mode," and be physically connected to the network to be considered a test station by NetBench. Test Station Coordinator One test station computer running NetBench (selected by the user) to synchronize the test and collect test station results. The Test Station Coordinator also acts as a test station. It performs the same tests as the other test stations and then reactivates as the coordinating test station at the completion of each NetBench test. WS.ID Each test station must a have local C: drive. Each test station must also have a file in the root of the local C: drive called WS.ID. This file must contain a unique three digit test station identifier in position 1 of the file (no leading spaces or line feeds before the identifier). Example: 001. For administrative purposes, it is a good idea to sequentially number each test station across the network. In addition, this file may also contain a seven-digit alpha-numeric Group identifier. If this is used, it must be separated from the test station ID by at least one space. Example: 001 GroupA1. Group identifiers are not case sensitive, but characters other than numeric and alpha must not be used. When a Batch Run File is used, Group identifiers are referenced from the batch file. Licensed Material (C) Ziff-Davis Publishing Company, L.P. Copyright (C) 1992, 1993 by Ziff-Davis Publishing Company, L.P. All rights reserved. Manual release date: August 1993 with Version 2.10 of NetBench