RAID, What is it and why do you care?


See also: Building Redundancy with a SATA RAID Array

RAID is an acronym for Redundant Array of Intelligent Disks. Sometimes the acronym is quoted as Redundant Array of Intelligent Devices or Redundant Array of Inexpensive Disks. In any case it is a group of hard drives arranged to optimize your storage to suit your individual needs. The optimization is either done with software (Software RAID) or hardware (Hardware RAID).

Hard drives configured in a RAID setup are called an array. There are many types of RAID arrays some of them that you may see referenced are RAID 0, RAID1, RAID 5, RAID 6, JBOD and then there are the combination arrays that sometimes are referred to a two digit number and some times as one number + another number. i.e. RAID 0+1 or RAID 10, RAID1+5 or RAID 15. Because different sources identify these configurations differently I will not attempt to identify all the possible configurations and all the possible acronyms for them. I will however try to explain what the most common types of RAID are, and how they can affect your small business.

RAID 1 provides hard drive redundancy. Two identical hard drives are installed, one is loaded with the operating system and all your programs and data, the other copies the first drive constantly. This is called a mirror and if one of the drives fails you can set the other drive to take over with no loss of data and no down time. The failed drive can be replaced later or even while the computer is running on some configurations (those would be the more expensive ones).

Mirroring is highly recommended for Small Business operations, you just can not risk losing your data or shutting your network down while you try to recover your data. The downside to mirroring is that the copy process slows the system down a bit. If you are running a Software RAID mirror, the slow down is going to be noticeable, but not unreasonable. Hardware mirroring is nearly undetectable.

RAID 0 provides high performance, commonly called Striping. Two (or more) identical drives are installed and the RAID controller configures them both as one drive of approximately twice the size. When the system reads from or writes to the new single drive, it actually reads or writes stripes of data across both drives at the same time. This reduces the time it takes to access or store the data. The downside to Striping is that if one drive fails, all the data is lost. (Bummer!)

A common compromise used to be RAID 10 which stripes two sets of two drives for speed, and then mirrors one striped drive to another. The end result is faster then a regular mirror, and slower then a regular striped drive. If any one drive fails, the mirrored stripe continues to function. The downside here is you need 4 or more drives, which drives the cost up.

RAID 5 was designed to resolve all the issues. RAID 5 stripes across three or more drives, in the process it writes an extra recovery stripe. If one drive fails, the other drives can rebuild the data from the recovery stripe information. Again this is often offered in a “Hot Swap” arrangement so the dead drive can be replaced, and rebuilt with out even rebooting the machine. Until recently RAID 5 was only available in high end SCSI setups, not commonly found in Small Business offices. The only downside to RAID 5 is that you lose about 1/3 of the drives total capacity to the recovery stripe. (A small price to pay!)

RAID 6 is a new configuration being touted by Intel and others as the next generation. At this point I will leave the advantages of RAID 6 for you to determine on your own if you are interested. Read more about it here.

JBOD is "Just a Bunch Of Disks". Basically you have a RAID controller and you are using it just to increase the number of drives you have. There is no real RAID function in JBOD in the traditional sense of the term.

The thing that makes RAID a point of interest at this time is the development of the SATA II standard and the introduction of inexpensive SATA RAID controllers built right on the motherboards. In the past this was not unheard of for RAID 0, 1 and 0+1 but RAID 5 was SCSI only. Today you can buy a motherboard for $200 or less with SATA II RAID 0, 1, and 5 (and all the combinations) built right on. Additionally you can buy SATA II controllers from companies like Adaptec which will allow you to run any SATA RAID configuration you want on pretty much any modern motherboard you want.

The real kicker here is that SATA drives are inexpensive. In the past a RAID configuration would include a 36GB SCSI hard drive that cost in the neighborhood of $500 each.

So take a five hard drive RAID 5 solution as an example: Five 36 GB hard drives @ $500=$2500 this would yield around 120GB of space. In addition a RAID 5 SCSI controller card would be required, these typically run $350 to $700 more. Server grade motherboards came with built in RAID 5 SCSI controllers, but the cost was easily $300 higher then a motherboard with out the controller. So the end cost here is between $2800 and $3200 for 120Gb of space (or a minimum of $23 per GB).

Now let’s look at the new SATA II RAID 5 configuration: You can buy a motherboard for under $150 with onboard RAID 5, or a SATA II RAID 5 controller for about $200 to $250. You can buy five 120 GB SATA II hard drives for something like $500. This would yield about 480 GB of space. The final cost then is between $550 and $750 for 480 GB of space ($1.15 per GB on the low end).Now let’s take a quick look at the cost of mirroring. If you have a Windows 2000 or newer server you can run a software mirror that is built into the OS, as I mentioned there will be a detectable slow down in the system. You can run a hardware mirror in your server, or even on a single desktop. The cost for two hard drives and a controller card would come in the neighborhood of $300 to $350 (at the time this is writeen SATA II RAID cards, without the RAID 5 function, are about $150, 120GB drives are about $100 each).

Consider what would happen tomorrow if you were working away and your hard drive failed. If you have been reading this webpage then you probably have a backup, but how long will you be down while you have the drive replaced, the software reinstalled and the data restored from backup? You DO have a backup, right?

Would you pay $300 to be certain you would never lose your data and not have to shut down the machine during a busy workday for because of a failure?

I would (already did).

Just FYI. here are a couple of website with more information on RAID. Remember, some details like the acronyms do not agree from one author to the next, but the concept and the execution are the same.

University of Massachusetts
Webopidia.com
Adaptec

-Steve Weigle