Real-World Scripting: Test Copy Throughput With a Script

As the administrator of a resource domain that services several thousand users, management has asked you to determine benchmarks for the transfer rate of files (i.e., copy throughput) between the domain servers and the lab computers. You've noticed a decrease in performance but haven't been able to diagnose conclusively which servers, hard disk arrays, and network links are having problems. You decide to develop a script that measures and logs copy throughput. Specifically, you want the script to

Meeting the Script Requirements
Although you could write the FileCopyTest script in Windows NT shell scripting, using Perl is better because Perl provides more accurate math functions than NT shell scripting. However, to use Perl, you must overcome two hurdles.

The first hurdle is how to make the script portable given the 100MB test file. One solution is to put the test file on a network server. However, with this solution, the script takes longer to execute because the source computer must copy the test file from the network server before the testing begins. A better solution is to include a section of code in the script that creates the test file on the source machine. At the end of the test, the user has the option of deleting the test file or leaving the test file in place for another test operation. As the code in Listing 1 shows, if the script detects an existing test file, the script skips the test-file creation code and proceeds directly to the testing code. Listing 1 is only an excerpt from FileCopyTest.pl. You can find the entire script, complete with comments, in the Code Library on the Win32 Scripting Journal Web site at http://www.win32scripting.com.

The second hurdle is that the Perl library must be on all the source computers that you want to test. In this instance, that number is in the thousands, so installing the Perl library on all of them would be too time-consuming. One solution is to place the Perl library on a server and have the script point to that library. However, this solution requires that you set up and maintain a share point just for testing copy throughput. Some companies try to avoid using formal shares (even hidden ones) for short-term administration purposes because these shares tend to become orphaned but never deleted for no one can remember who created them or why they were created. A better solution is to convert your completed Perl script into an .exe file with a command-line conversion program, such as DynamicSTATE's Perl2Exe or ActiveState's PerlApp. If you convert FileCopyTest.pl to an executable, you don't need to install the Perl library on all the source computers. In addition, with an executable, you can distribute the code on a disk yet protect the code from accidental or intentional modification. The Code Library on the Win32 Scripting Journal Web site includes the executable FileCopyTest.exe.

You can download free trial versions of Perl2Exe and PerlApp. Perl2Exe is available at http://www.dynamicstate.com. PerlApp, which is part of ActiveState's Perl Dev Kit, is available at http://www.activestate.com/pdk. If you plan to regularly distribute code to remote machines or machines on which users haven't installed Perl, you might consider purchasing a command-line conversion program.

Preparing, Converting, and Running the Script
I wrote and tested FileCopyTest.pl on servers and workstations running NT 4.0 with Service Pack 5 (SP5) and on workstations running Windows 2000 Professional (Win2K Pro) Professional pre-release. Here are the steps to prepare, convert, and run this script for your system:

Viewing the Results
After you run FileCopyTest.exe, the results will appear both on screen and in the log file. Figure 1 contains an example of a resulting log file.