As we've outlined earlier, one of the principal reasons for our delay in upgrading our testbed was the wait for Windows 2000. When we initially launched two years ago, Microsoft was touting Windows 2000 as the successor to both Windows NT 4.0 and the upcoming Windows 98. Thus, we longingly looked forward to the nirvana of a single operating system whose arrival would signal the end of our lengthy dual-OS procedures.
Things didn't quite work out that way. Microsoft eventually backpedaled, deciding that "Windows 2000 Consumer" would still be based on the Windows 9x kernel incorporating only minor changes. The problem of dual operating systems yet remains. This time, however, there are some twists.
When every indication confirmed that Windows 2000 Professional would indeed meet its February 17th launch, we rapidly moved forward to acquire and assemble the hardware that would create our new testbed. Just last month we were assuming that we'd launch the new tesbed with Win2k Pro and Win98 SE as its two operating systems. MS threw us a loop, however, when they announced that Win2k Consumer, dubbed Windows Millennium Edition (ME), would ship on May 26th.
About three months after StorageReview.com's initial launch, Microsoft released Windows 98. Over the past year and half, we've accumulated quite a few protests over our use of Windows 95 (OSR2.1) rather than Windows 98 (and eventually Win98 SE) on our testbed. We were caught in a paradox, so to speak. On the one hand a control article we published in July of 98 showed there was little change in disk performance between Win95 and Win98. On the other, such a negligible difference to many meant that there should be little problem moving to Win98 while keeping our growing database valid. We judged the latter option to be the greater of two evils and decided to stick with Windows 95.
We're in the same quandary now. We're launching our next-generation testbed at the same time that Microsoft is promising a revision of their consumer-oriented operating system within a few months. This time, however, we've decided to hold off Windows 9x testing until Windows ME arrives. Until then, we'll present results and reviews using Win2k.
The Installation of Windows 2000 Professional:
Our testbed's installation of Windows 2000 Pro is quite straightforward. The Adaptec ASC-29160's bios in conjunction with the Win 2k CD allowed us to boot directly into the installation procedure. We created a 5 GB partition, formatted with NTFS, on our Quantum Fireball lct10 boot drive. In all other ways, the Win2k installation was purely "stock," saving only the customization of network settings to disable File and Printer sharing.
Upon booting up the installed OS, our first task was to update the stock Win2k Voodoo3 driver with 3dfx's official v1.00.00 release. After doing so, we set the resolution to 1024x768, stuck with standard small fonts, and upped the refresh rate to 85 Hz. Using RegEdit, we disabled CD-ROM AutoPlay. Swapfile settings were left as is, which on a system with 128 megs of ram, ranges from 192-384 megs in size on the C drive. Using Windows Update, we patched the installation with the following updates:
- Windows 2000 Critical Update, February 17, 2000
- Windows 2000 Compatibility Updates
- Windows 2000 Update, February 17, 2000
- High Encryption Pack for Windows 2000
The next step was the installation of several benchmarks: ZD's WinBench 99 v1.1, Intel's IOMeter v1999.10.20, TestaCD Lab's HDTach v2.61, Adaptec's ThreadMark v2.0, and Atto Express Tools v1.44. Of these, however, WinBench 99 and IOMeter are the two that will be used regularly to assess drives.
After installing the benchmarks, each was loaded up (to initialize and bypass first-time welcome screens) and exited. We then ran RegClean, a utility that quickly scrubs the Windows registry of excess garbage. Finally, the boot drive was defragmented using Win2k's built-in utility.
ATA drives are installed into the testbed as the master (and only) device on the Promise Ultra66's secondary channel. SCSI drives are installed into the testbed as the ID 0 unit on the LVD segment of the Adaptec ASC-29160.
The above represents a stable, unchanging platform under which our ATA and SCSI drive tests are conducted. In a moment we'll take a look at the benchmarks we've deployed, but first, a look at why we use benchmarks in the first place.