In my previous article on non-standard RAID levels, I talked a bit about RAID 1E, which is a RAID level that provides RAID-10-like functionality but with an odd number of disks. Although disks are pretty cheap these days, you never know when you might need to save a few bucks on a project!
In this article, I'll provide a look at two other non-standard and closely related RAID levels -- RAID 5E and RAID 5EE.
With an E that stands for Enhanced, RAID 5E is a RAID 5 array with a hot spare drive that is actively used in the array operations. In a traditional RAID 5 configuration with a hot spare, the hot spare drive sits next to the array waiting for a drive to fail, at which point the hot spare is made available and the array rebuilds the data set with the new hardware. There are some advantages to this operational method:
- You know for a fact that the drive that would have been used as a hot spare is in working order.
- There is an additional drive included in the array, thus further distributing the array's I/O load. More spindles equals better performance in most cases. RAID 5E can perform better than typical RAID 5.
There are a few disadvantages associated with RAID 5E as well:
- There is not wide controller support for RAID 5E.
- A hot spare drive cannot be shared between arrays.
- Rebuilds can be slow.
The capacity of a RAID 5E array is exactly the same as the capacity of a RAID 5 array that contains a hot spare. In such a scenario, you would "lose" two disks' worth of capacity -- one disk's worth for parity and another for the hot spare. Due to this fact, RAID 5E requires that you use a minimum of four drives, and up to eight or 16 drives can be supported in a single array, depending on the controller. The main difference between RAID 5 and RAID 5E is that the drive that would have been used as a hot spare in RAID 5 cannot be shared with another RAID 5 array; so that could affect the total amount of storage overhead if you have multiple RAID 5 arrays on your system. Figure A gives you a look at a RAID 5E array consisting of five drives. Take note that the "Empty" space in this figure is shown at the end of the array.
|A RAID 5E array with five drives|
When a drive in a RAID 5E array fails, the data that was on the failed drive is rebuilt into the empty space at the end of the array, as shown in Figure B. When the failed drive is replaced, the array is once again expanded to return the array to the original state.
|A RAID 5E array that has been rebuilt into the hot spare space|
RAID 5EE is very similar to RAID 5E with one key difference -- the hot spare's capacity is integrated into the stripe set. In contrast, under RAID 5E, all of the empty space is housed at the end of the array. As a result of interleaving empty space throughout the array, RAID 5EE enjoys a faster rebuild time than is possible under RAID 5E.
RAID 5EE has all of the same pros as RAID 5E but enjoys a faster rebuild time than either RAID 5 or RAID 5E. On the cons side, RAID 5EE has the same cons as RAID 5E, with the main negative point being that not a lot of controllers support the RAID level yet. I suspect that this will change over time, though.
As is the case with RAID 5E, RAID 5EE requires a minimum of four drives and supports up to eight or 16 drives in an array, depending on the controller. Figure C shows a sample of a RAID 5EE array with the hot spare space interleaved throughout the array.
|A RAID 5EE array with five drives|
When a drive fails, as shown in Figure D, the empty slots are filled up with data from the failed drive.
In my previous article on RAID 1E, some readers mentioned that RAID 1E simply doesn’t seem like a good alternative to RAID 10, particularly since hard drives are so cheap these days. I happen to agree that there would need to be a seriously special case to consider RAID 1E. With regard to RAID 5E and RAID 5EE, however, I can see a very positive upside with regard to performance, especially for organizations that are already using or are considering RAID 5.