RAID groups can seem more complex than they really are. That's because there are approximately 12 different types or variations of RAID available today and even more are defined. For the vast majority of small-midsized businesses (SMBs), there are only six RAID levels that really matter. Choosing the right RAID level depends on the application data type, the criticality of that data and the number of users. It first requires a small...
bit of knowledge as to what each of these six RAID levels can do.
RAID 0 splits or stripes the data across all the drives in the RAID group. The benefit of RAID 0 is higher data throughput. The downside is that no redundancy makes performance good, but any disk fault or failure results in complete data loss.
You should choose RAID 0 when increased storage performance is very important, budgets are highly constrained and potential data loss isn't an issue. Some good application data examples include temporary photo editing or video editing files.
RAID 1 synchronously mirrors all the data from each drive to an exact duplicate drive. No data is lost if either drive has a fault or failure. The advantage of using RAID 1 is improved multi-user read performance, because either disk can be read at the same time. The disadvantage is that the storage drive cost per usable byte of storage is twice as much because two drives are needed to store the same data.
Choose RAID 1 for an application requiring a belt-and-suspenders approach (i.e., you can't afford the possibility of application data loss or disruption), plus high-performance random reads. A good example is a read-only database in a retail brick-and-mortar store. RAID 1 is also a good choice for entry-level systems where only two drives are available, such as in a small file server.
RAID 10 (i.e., RAID 1+0 and RAID 0+1)
RAID 10 is the combination of RAID 0 and RAID 1. The advantage of using RAID 10 is the redundancy of RAID 1 with the performance of RAID 0. System performance during a drive rebuild is also noticeably higher than parity-based RAID levels (RAID 5 and RAID 6). This is because data doesn't require regeneration from parity information because it's just copied from the other mirrored drive. The disadvantage is the much higher cost --usually 60% to 80% more than parity RAID levels.
There are two variations of RAID 10. The first is RAID 0+1, which stripes the data across multiple disks, and then the striped disk sets are mirrored to an identical group of disks. The second is RAID level 1+0, which mirrors the data where the mirrors are striped.
You should choose RAID 10 for applications requiring the higher performance of RAID 0 and the unparalleled data protection of RAID 1. Online transactional databases often fit that bill.
RAID 5 is designed to provide RAID 0 performance with more economical redundancy, and is the most common RAID level in the majority of organizations. It does this by striping block data across several drives and distributing parity among the drives. No single disk is devoted to parity. The benefits of using RAID 5 are overlapped read and write operations (more efficient use of disk drives) speeding up small writes in a multiprocessor system, providing greater usable storage than RAID 1 or 10 (the redundancy storage penalty is approximately 20% instead of 50%). Data protection comes from parity information that's used to reconstruct data if a drive in the RAID group has a fault or fails. The downsides include a minimum of three -- and usually five -- disks per RAID group, significantly lower performance of the storage system while a drive is being rebuilt and the potential of total RAID group data loss if a second drive fails while the first is being rebuilt. Additionally, read performance tends to be lower than other RAID types because parity data is distributed on each drive.
You should choose RAID 5 for the vast majority of applications as long as the disk drives aren't high-capacity SATA drives. SATA drives have lower duty cycles than SAS or Fibre Channel drives and have shorter MTBF rates. And because SATA drives are very high capacity (500 GB to 1000 GB), rebuild times are very long, with controller performance degraded as well. High-capacity SATA drives also increase the probability of a second disk fault or failure, resulting in complete data loss.
RAID 6 is similar to RAID 5, and includes a second parity scheme distributed across the drives of the RAID group. The advantage of using RAID 6 is that the second parity protects against data loss should a second drive fault or fail in the RAID group. This makes the high-capacity SATA drives more viable and economical than RAID 1 or RAID 10. The disadvantage of using RAID 6 comes from a much lower storage system performance if there are two drives rebuilds going on simultaneously, often below 20%.
You should choose RAID 6 for high-capacity SATA drives and applications that can tolerate reduced performance at times. Some application examples include archiving and non-dynamic multimedia data such as JPEGs and streaming video.
RAID 50 (also known as RAID 5+0 and RAID 0+5)
RAID 50 is the combination of RAID 0 and RAID 5. It takes RAID 5 RAID groups and stripes them like RAID 0, which increases performance. The advantage of RAID 50 is increased performance over standard RAID 5. The downside is higher costs and lower usable capacity.
There are two variations of RAID 50. RAID 0+5, which utilizes multiple RAID 5 sets striped in a single RAID group. This increases the RAID group reliability by making it capable of tolerating a disk fault or failure in either or both of the RAID 5 sets without data loss. RAID 5+0 is the more common form of RAID 50, which takes RAID 5 groups and stripes them like RAID 0. You should choose RAID 50 for applications where the economics of RAID 5 are important but where there's a requirement for better performance.
About this author: Marc Staimer is President and CDS of Dragon Slayer Consulting in Beaverton, OR. He's widely known as one of the leading storage market analysts in the network storage and storage management industries.