A recent article by Computing Technology Review’s Tom Joyce discusses the fact that with every major advancement in enterprise computing comes new set of performance management challenges. This was and is certainly true of RAID technology–even though rumors have circulated in some circles for years that RAID is a performance solution and doesn’t need defragmentation, analysis as well as user testimony has certainly shown otherwise.
RAID (which originally meant Redundant Array of Inexpensive Disks and now means Redundant Array of Independent Disks) utilizes any of several technologies to create a virtual volume which, in fact, is made up of numerous disks. Data written to the RAID set is written redundantly so that if one of the disks should fail, all of the data is still available, many times with no visible interference to the user. This means greatly reduced or no downtime for vital applications such as databases. The operating system “sees” the RAID set as one volume; data is written physically to the disks through the RAID controller which also create a copy of each data block.
RAID technologies also serve to improve performance; since the I/O load is across several disks instead of just one, retrieval is faster. But the myth that fragmentation does not affect RAID arrays is truly that: a myth. As an application reads and writes to this virtual environment (creating new files, extending existing ones, as well as deleting others) the files become fragmented. Because of this fact, fragmentation on this logical drive will have a substantial negative performance effect. When an I/O request is processed by the file system, there are a number of attributes that must be checked which cost valuable system time. If an application has to issue multiple “unnecessary” I/O requests, as in the case of fragmentation, not only is the processor kept busier than needed, but once the I/O request has been issued, the RAID hardware/software must process it and determine to which physical member to direct the I/O request.
A defragmenter sees the RAID environment just as the file system does. That is, it defragments the logical drive, improving the speed and performance of a RAID environment by eliminating wasteful and unnecessary I/Os from being issued by the file system. This occurs because the file system sees the files and free space as being more contiguous. The file system will spend less time checking file attributes which means more processor time can be dedicated to doing real useful work for the user and application.
With today’s enormous disk capacities and file sizes, combined with higher-than-ever access demand on RAID sets through servers, attention must also be paid not only to the fact of defragmentation, but the defragmentation technology itself. Scheduled defragmentation, once “state of the art”, is now leaving fragmentation behind after defragmenter runs. The true solution in high-demand server environments is a new breed of defragmenter which requires no scheduling but runs invisibly in the background, whenever system resources are available, and keeps pace with today’s frantic fragmentation rates.