Defragmentation Boosts NT Performance After All
In the debate over the impact of disk fragmentation on system performance, the faction that claims that defragmentation software boosts system performance on Windows NT, both on the workstation and server, now has some convincing evidence on its side.
In the debate over the impact of disk fragmentation on system performance, the faction that claims that defragmentation software boosts system performance on Windows NT, both on the workstation and server, now has some convincing evidence on its side.
The National Software Testing Lab (NSTL, www.nstl.com) recently conducted independent performance testing on defragmentation software that showed Windows NT Server 4.0 performance increased by 56 percent and Windows NT Workstation 4.0 recorded gains of 81 percent.
Utility vendors -- such as Executive Software Inc. (www.execsoft.com), Symantec Corp. (www.symantec.com) and Raxco Software (www.raxco.com) -- have historically claimed significant performance boosts, but they lacked the hard evidence that these tests provide.
The detractors had argued the fragmentation problem that plagued DOS was minimized by Windows NT technologies such as seek optimization, reordered I/O requests and file-based cache.
Windows NT 4.0 and earlier iterations lack a built-in defragmentation tool, but that will change with Microsoft Corp.’s forthcoming release. Technology from Executive Software’s Diskeeper 4.0 was added to the third beta of Windows 2000 Professional, Server and Advanced Server. Datacenter Server will include the tool as well. Diskeeper 4.0 is the software package NSTL used in its tests.
Richard Kurczewski, the defragmentation test project manager at NSTL, says the results are not tied exclusively to the Diskeeper software.
Andy Staffer, director of research at Executive Software, agrees. "What we’re dealing with here is fragmentation, so anybody can get similar benefits with a different software package if the theory behind it is defragmentation."
NSTL tested two of the most common real-world Windows NT workstation and server configurations, as determined through a survey of 6,000 systems managers.
A Pentium II 266-MHz workstation with 96 MB of RAM and a 2 GB IDE hard drive running Outlook and Excel showed a performance leap of 74.5 percent with a defragmented drive.
A more powerful machine scored higher on the same test. When NSTL ran a Pentium II 400-MHz workstation with 228 MB RAM and a 4.2 GB hard drive, the improvement was 80.6 percent.
On the server side, a dual Pentium Pro 200-MHz machine with 128 MB of memory and five 4-GB SCSI hard drives running RAID 5, Exchange Server and SQL Server 7.0, attained an increase of about 32.4 percent on a defragmented drive.
The boost was 56.1 percent on the other server, a Pentium Pro 200-MHz machine with 64 MB of RAM, two 4-GB SCSI hard drives running Exchange and SQL 7.0.
"Even without fragmentation, the drives on the server side already move quite a bit faster than on the workstation side, so there is less opportunity for performance increase," NSTL’s Kurczewski says. "Still, with defragmentation, performance on the servers increased dramatically, more than enough to make it worthwhile."
Disk drives pose the biggest system bottleneck. CPU speeds are disproportionately faster, so drives can’t keep up. Although the higher-end systems performed better in NSTL’s tests, defragmenting disk drives is a viable alternative to throwing hardware at the system to make it faster.
When Windows 2000 is released, with Diskeeper technology included, the performance enhancement the defragmentation technology will offer is expected to be on par with NSTL’s results for Windows NT 4.0.
The version of Diskeeper included in Windows 2000 is a lighter version than the shrink-wrapped box. The Windows 2000 version shares the Diskeeper 4.0 defragmentation engine but lacks the scheduling capabilities, server operations and priority settings. A Diskeeper for Windows NT 4.0 has been available for free at Executive Software’s Web site for two years. It has been enhanced to include necessary Windows 2000 functionality, such as support for FAT 32 and 64-bit code.
Defragmenting Other Microsoft Apps
The National Software Testing Lab also tested defragmentation software on four Microsoft applications: Excel, SQL Server 7.0, Outlook 98 and Exchange 5.5.
- Excel ran 135.2 percent faster when the drive was defragmented.
- A SQL Server 7.0 Table Creation Benchmark ran 32.4 percent faster with the drive defragmented, and a Query Benchmark was 23.7 percent faster with defragmentation.
- A Workstation running Outlook opened 50 messages 33.3 percent faster when defragmented compared with a fragmented system. The same system searched all messages for a particular word 25.5 percent faster than a fragmented system, and searched all messages for attachments 55.6 percent faster than a fragmented system.
- The defragmented server opened 50 messages 5.1 percent faster than a fragmented system. The same system searched all messages for a particular word 31.2 percent faster than a fragmented system, and searched all messages for attachments 17.6 percent faster than a fragmented system.