Some time ago, Microsoft completed a performance test against then-rival Novell. The results were disputed -- by Novell, independent experts, the press, network administrators and consultants with extensive experience with multiple operating systems.
Super-secret Windows Registry hacks that caused constant crashes were involved, but part of Microsoft's settup was instructive. The Microsoft engineers separated the operating system and paging files into separate volumes, then split the work area into four separate data volumes. They cited performance bottlenecks in the NTFS file system journal as the reason for using mulitiple volumes instead of one single volume.
A journaled file system must handle journal writes in a single-threaded fashion; a large number of simultaneous writes could theoretically bog down the system. By creating multiple volumes on the array, the Microsoft engineers in effect multiplied the number of journal queues, eliminating this as a potential bottleneck.
While system performance is much higher today, the lesson is still applicable with today's astoundingly high disk capacities. Splitting a high-access server's data into several discrete volumes (at the time of install, rather than with a design retrofit such as DiskPart), concatenated on a single array, can potentially improve system performance. This also has the added benefit of reducing downtime if a single volume needs to be repaired or restored.
Rating this tip below will let
us know how useful it was to you. And if you have a Windows tip, timesaver or workaround to share, submit it to our tip contest. You could win a $150 gift certificate.
This was first published in June 2007