Can Heterogeneous RAID Arrays Work?


While the integration of basic RAID support into chipset core logic has made it very easy to create RAID 0 and RAID 1 arrays using up to six hard drives (almost every AMD, Nvidia or Intel chipset supports RAID in the upper mainstream and higher product categories), it is always important to have a close look at the basic question of whether or not a RAID array really makes sense. In this context, I’d like to point to the difference between redundant data storage using RAID technology, and making regular backups to protect data from intended or accidental manipulation as well as so-called disasters such as power surges, fires, flooding, earth quakes and other disasters. RAID, which stands for redundant array of independent drives, is by no means a replacement for backups, as it can only be used to speed up your storage or to store existing data in a redundant manner. The primary objective behind the use of RAID is to store data redundantly in an effort to make the solution immune to failure of individual hard drives.

RAID does not, however, protect your data against catastrophes. Users working with data on RAID arrays may very well modify or delete data accidentally, or one of the disasters mentioned above can destroy all your digital assets within seconds. For these reasons, it is important to backup your important data to other media on a regular basis and to consider redundant storage as a second step. Once your backup strategy is in place (and tested, because a backup is only helpful if the recovery works), it’s time to consider RAID storage.

Every system you need to be available 24/7 should be based on a fully redundant partition. RAID 1 has turned out to be the best solution in most cases, as the content of one hard drive is simply mirrored onto a second drive in real time. If one drive fails, your system will still be operational. Different from all the other RAID modes, RAID 1 does not have any impact on performance. RAID 0 cannot be considered a true RAID mode, as it distributes data across all available hard drives; hence it does not offer any data redundancy. The more sophisticated RAID modes 5 and 6 accommodate multiple hard drives and hence allows for the creation of really large arrays while offering single and double redundancy. With RAID 5, one drive is used to store redundancy data while RAID 6 creates two sets of redundancy data. Both provide more net storage capacity than in a simple RAID 1, as you get the total capacity of all RAID drives minus the capacity of one drive (RAID 5) or two drives (RAID 6). But RAID 5 and 6 require a minimum of three and four drives, respectively, to work, and the parity calculation requires a significant amount of processing power. If you want a RAID 5/6 to perform really well you will have to go for a fast RAID controller and more than the minimum amount of drives, as a RAID 5 with three or four drives is usually outperformed by a single hard drive.

The best approach is to ask you: Where do I need data to be stored redundantly and which RAID level makes the most sense from an investment and performance standpoint? If system downtime is an unviable scenario, then you will have to go for a RAID solution. In other cases, you can still consider single hard drives with a replacement available next door combined with a tight backup strategy.

  • Typically, I'd say, run RAID-1. When it's time to upgrade, get two more identical drives, plug-em in and set up another mirror. Then, for Windows, delete your system hdd from the device manager and reboot to a Norton Ghost CD (assuming it supports your raid controller). Clone the partitions to new mirror. Power off, remove old mirror, boot to a new mirror. It will most likely re-detect the new mirror and request a re-boot, after that you're good to go on. Old drives can be wiped and sold on e-bay. Or turned into a backup volume.