RAID: What is it and what do I really want?
To start, when you think of RAID, do not think of the spray can you use for ants. RAID, in this case, stands for Redundant Array of Independent (formerly the I stood for Inexpensive) Disks. In other words, you have multiple hard drives in your computer that work together in such a way that the computer doesn't see anything more than a single hard drive. So why would you want this? Because, as the acronym implies, it gives you fault tolerance (if one of the drives in the array fails, you still have access to all your data) and peace of mind (the fault tolerance means that your data is not as susceptible to loss as it would be on a single drive).
Okay, so now you know that you want RAID. What more do you need to know? This is where it gets more confusing and technical. I'll try to make it easy to understand for you.
There are many different configurations to RAID, called RAID Levels. Each level means that the array is used in a different fashion to accomplish different things.
The first level that is not really RAID but is still considered part of the level hierarchy is RAID Level 0, also known as striping. What this means is that the data is striped across the various hard drives in the array so that you gain the performance increase of having multiple drives serve up your data at the same time and the security of knowing that if any one of the drives gets stolen, the person with that drive will have nothing useful. All the drives must be present, installed in the correct order, and have the array configured the exact same way on the "new" system as it was on the original system so that the data can be accessed. The disadvantage to RAID Level 0, though, and the reason that it is not really a RAID is that there is no redundancy and, therefore, no fault tolerance. In other words, if one of the drives fails, you lose all your data. For this reason, we do not recommend RAID Level 0 unless all you need is performance and would not be inconvenienced by reinstalling your applications and recreating your data (assuming you have no backups) when a drive failure occurs. Gamers tend to use RAID Level 0 to get higher frame rates in the games.
Next is RAID Level 1, also known as mirroring. This has two exact drives (same make and model) that have the same data written to them at the same time. If one of the drives fails, the computer lets you know but you keep right on working away as if nothing has happened because you still have all your data available to you. Once you get a replacement for the failed drive, all you need to do is install it and turn the computer back on. The RAID array will be rebuilt automatically. In time, both drives will have all the data and the array will be back to a "normal" state. This method is used by smaller file servers commonly and was the method of choice for most small business file servers throughout the 1980s and 1990s. It's not used as commonly today because the price of hard drives has plummeted so far that utilizing other RAID options is much more affordable than in previous years. The disadvantage to this array is that performance is not increased as it is with RAID Level 0 and, as will be discussed next, RAID Level 5. In fact, performance is sometimes slightly degraded because the same hard drive controller is trying to write to both drives at the same time. Thus, if you use this, we recommend not doing so on a machine that is going to experience high utilization.
Though there are RAID Levels 2, 3, and 4, they have not been used anywhere that I've seen in my experience of 15 years in the networking field, so we are going to skip over them and go right to RAID Level 5, which is Striping with Distributed Parity. RAID Level 5 utilizes a minimum of three disks but the most common configuration is five and distributes the stripe (frequently 64K in size) across all drives with one of the drives in the array holding the parity data. The drive that holds the parity information for each successive stripe rotates. Thus, in a five-disk array, for the first stripe, the fifth drive would have the parity. In the next, it would be the first drive. The next would be the second drive and so forth. This insures that there is no data loss because the missing data can be calculated from the distributed parity across the remaining drives and, thus, the user (or, in the case of a file server, users) never realize there has been a failure. Only the one who operates at the console of the computer realizes it and knows to swap out the failed drive. Just like with the other RAID Levels, the replaced drive is rebuilt automatically by the system. This is the most common RAID configuration used in file servers today.
Just like with RAID Level 0, this has the advantages of file access performance and security. However, since it also has the parity feature, this does not have the disadvantage of RAID Level 0, which is the potential loss of all data due to the failure of only one drive. Therefore, if someone is considering RAID Level 0, we strongly urge reconsideration and highly recommend RAID Level 5 instead.
The final basic RAID configuration is RAID Level 6, which is not commonly used but may be seen in some high-priority environments. RAID Level 6 is a redundant RAID Level 5, you could say. It is Striping with Dual Distributed Parity. In other words, instead of one drive with parity information, the array has two. This also increases the minimum number of drives from three to four. The advantage to this is that if the array has not one but two drives fail at the same time, then all the data is still preserved and accessible. With RAID Level 5, a dual-drive failure occurrence would cause a loss of all data.
On occasion, one may see RAID Levels such as 01, 10, 50, or 51 (or they can be expressed as 0+1, 1+0, 5+0, or 5+1). These are nothing more than the combination (sometimes referred to as Nested or Hybrid Levels) of RAID Levels 0, 1, and 5 in various combinations. If one is going to use this additional level of complexity, we recommend RAID Level 51. This means that the primary level is a striped array with distributed parity but that it is also mirrored. The minimum number of drives for this configuration is six. As it must be the same number of drives in both sides of the mirror, the number of drives involved will always be an even number.
As you can see, there are many options to consider based on the way the computer will be used. If you feel that putting a RAID array into operation is something you need, let us know and we will be happy to discuss the options with you and give you a quotation to fit your needs.