When you buy through affiliate links in our content, we may earn a commission at no extra cost to you. Learn What is RAID?
RAID, or Redundant Array of Independent Disks, is the combining of multiple physical hard drives to appear and act as a single virtual/logical disk. There are quite a few ways to do this, called “levels.”
Techopedia Explains
Some RAID levels, such as RAID 1 or RAID 5, are commonly deployed, though there’s constant debate over which level works best and under which circumstances. RAID levels such as 2, 3, and 4 are rarely implemented for various reasons, and 7 is a proprietary implementation.
If you want to secure your data against hardware failure, a RAID level that offers redundancy via mirroring or parity is a must. It’s a sad day when a solo drive or striped/no parity RAID 0 array goes belly up. Recovery is generally possible but very expensive. RAID with mirroring or parity also allows you to keep working until the faulty drive is replaced. That can save you scads of money.
RAID levels that increase sustained throughput (random access generally suffers slightly in any RAID setup) through striping or otherwise distributing data across multiple disks can also be of great benefit. They allow you to leverage older, slower storage technology such as 250MBps hard drives or even 550MBps SATA SSDs into storage that can stream large amounts of data far faster. We’ve seen even small 4-bay NAS boxes exceed 1GBps with striping.
Hardware vs. Software RAID
Hardware RAID is implemented through dedicated controllers (HBAs or Host Bus Adapters), while software RAID is implemented via an operating system using normal CPUs. You might think most NAS boxes are hardware RAID. In actuality, it’s more often software RAID implemented by the OS that also provides the other functionality. That’s why today’s NAS boxes are so versatile.
The operating system is generally Linux with end-user boxes, and FreeBSD or other UNIX variants running file systems such as ZFS or OpenZFS come into play when you reach the larger business or enterprise level.
Software RAID is easier to implement, update, and improve, but dedicated hardware controllers generally have an advantage in speed, ease of implementation, and general focus on tasks.
RAID Levels
Here, we describe the various RAID levels, from basic to not-so-basic, the slow to the fast, and the safe to not-so-safe.
Note that usable capacity shown in each callout is the percentage of the raw disk capacity that remains available after the RAID level is implemented. For example, two 12TB hard drives in RAID 1 offer 24TB of raw capacity but only 12TB (50%) usable capacity.
JBOD (Just a Bunch of Disks)
Though not really a RAID level, JBOD is commonly offered as an option with multi-drive storage boxes. All it means is that the drives appear as one logical unit, and data cascades down the line. When the first disk is full, data is written to the next, and so on.
It’s a nice way to feel you aren’t going to run out of space but offers no redundancy or increased performance.
- Pro: 100% usable capacity
- Con: Decreased safety, no performance improvement
- Requirements: 2 or more drives, controller, or software
RAID 0 (Disk Striping)
RAID 0 distributes data evenly across disks for drastically improved sustained throughput but without the safety benefits (or overhead) of writing parity information or mirroring.
RAID 0 is dangerous because the failure of one drive brings the entire array down. The more disks in the array, the more failure points. This level, unless combined with other RAID levels, should only be used for scratch operations or, if important data is involved, continually backed up. Recovery is usually possible but very expensive.
- Pro: Increased sustained throughput, 100% usable capacity
- Con: Decreased safety
- Requirements: 2 or more drives, controller, or software
RAID 1 (Disk Mirroring)
RAID 1 is creating an exact copy of one disk on another. Basically, data is written to all drives at the same time. You can have as many mirrors as you want, but typically drives are paired. Should one fail, you still have a functional disk to work off of while the failed disk is replaced and the array rebuilt (the existing data copied to the new drive).
RAID 1 mirrors are often combined under other RAID levels (RAID 10, etc.) for increased performance while maintaining redundancy. The rather large downside to RAID 0 is that you lose half of your raw capacity.
- Pro: Increased data safety
- Con: No increase in performance, 50% usable capacity
- Requirements: 2 or more drives, controller, or software
RAID 1E (Striped Mirroring)
RAID 1E is striping a duplicate (mirrored) block of data to a different drive in a set of three or more. It’s not the parity info you’ll find in levels such as 5, but an actual duplicate/mirror of the data. The data block and the mirror/duplicate block can be on any of the drives as long as it’s not the same one. Like plain RAID 1, usable capacity is half of the total capacity.
- Pro: Increased sustained throughput and data safety
- Con: Only 50% usable capacity
- Requirements: 3 or more drives, controller, or software
RAID 2 (Bit striping with Parity and Error Correction)
In practice, RAID 2 has proven difficult and expensive to implement. A controller spreads data at the bit level across multiple hard drives with Hamming (linear error correction) code used to correct errors and provide parity bits.
Error Correction Code (ECC) on modern hard drives has largely removed the need for Hamming, but there is still a major technical hurdle in synchronizing the spindles for single-bit operations. Also, as all spindles are in play for every write, only one write operation at a time is possible. This makes for very fast sustained throughput but lousy multi-transactional (random) performance.
- Pro: Increased sustained throughput and data safety
- Con: Spindles must be synced, decreased random performance, 66% usable capacity
- Requirements: 2 or more drives, controller, or software
RAID 3 (Byte Striping with a Single Parity Disk )
Like RAID 2, RAID 3 is another level that sounds like it will provide a nice performance boost: bytes striped across multiple disks with all the parity info stored on a separate disk. But again, synchronized spindles were required, and performance is great for sequential reads and writes but weak for multi-tasking read and write operations.
- Pro: Increased sustained throughput, greater safety
- Con: Spindles must be synced, decreased random performance, 66% usable capacity
- Requirements: 3 or more drives, controller, or software
RAID 4 (Block Striping with a Single Parity Disk)
RAID 4 is much like RAID 5 in that it strips data at the block level across multiple disks (at least three). However, unlike RAID 5, which distributes parity info across all the disks in the array, RAID 4 parity is written to a single disk. The one pipe for parity can be a bottleneck under heavy loads.
- Pro: Increased sustained throughput, greater safety
- Con: Spindles must be synced, decreased random performance, 66% usable capacity
- Requirements: 3 or more drives, controller, or software
RAID 5 (Block Striping with Distributed Parity)
RAID 5 requires at least three drives and is similar to RAID 0 in that it strips (distributes evenly) data at the block level across multiple disks. However, unlike its life-on-the-edge cousin, it also strips parity info across all the drives in the array for data redundancy.
Parity blocks never reside on the disk that contains the data they protect. RAID 5 is the lowest RAID level to use distributed parity. RAID 5 can withstand the loss of one drive.
- Pro: Increased sustained throughput, greater safety
- Con: Nothing to speak of, 66% usable capacity
- Requirements: 3 or more drives, controller, or software
RAID 5E (RAID 5 with Distributed Hot Spare)
The E in RAID 5E stands for extended or enhanced, depending on who you talk to. Extended/enhanced, as in an extra disk, is included in the array as a hot spare, but that actually functions as part of the array. The spare capacity represented by the extra disk is spread over all the disks in the array in large contiguous areas at the end of the data on each disk.
The minimum number of disks is four in 5E, but otherwise, it’s RAID 5 striping with distributed parity.
- Pro: Increased sustained throughput, excellent safety
- Con: Long rebuild times, 50% usable capacity
- Requirements: 4 or more drives, controller, or software
RAID 5EE (RAID 5 with Distributed/Interleaved Hot Spare)
RAID 5EE is The same as 5E, but the empty space that the “hot spare” represents is interleaved with the rest of the data rather than in large contiguous blocks at the end of the data on each drive. Recovery time is said to be faster than with RAID 5E.
- Pro: Increased sustained throughput, excellent safety
- Con: Long rebuild times, 66% usable capacity
- Requirements: 4 or more drives, controller, or software
RAID 6 (Block Striping with Double Parity)
RAID 6 is a close cousin to RAID 5 in that data and parity info are striped across all the disks in the array. However, RAID 6 distributes twice the parity info–two blocks for every block of data, as opposed to RAID 5’s one. The original data block, parity block one, and parity block two all reside separately.
RAID 6 requires a minimum of 4 drives and can withstand the loss of two drives and keep on ticking. Normal RAID 5 can tolerate only a single drive failure.
- Pro: Increased sustained throughput, best safety
- Con: Long rebuild times, 50% usable capacity
- Requirements: 4 or more drives, controller, or software
RAID 7 (Proprietary)
RAID 7 isn’t one of the standard RAID levels but a proprietary solution integrating a CPU and controller offered by Storage Computer Corporation (now defunct). RAID 7 utilizes RAID 3 and RAID 4 features while adding the benefits of caching.
RAID 10 (RAID 1 + 0)
RAID 10 is a combination of RAID 1 mirroring and RAID 0 striping. It’s also referred to as 1 + 0, and we’ve also seen it as 0 + 1. Data is striped in RAID 0 fashion across one or more RAID 1 mirrors.
RAID 10 can withstand the loss of two drives as long as they’re not both in the same mirrored pair.
- Pro: Increased sustained throughput, very good safety
- Con: Long rebuild times, 50% usable capacity
- Requirements: 4 or more drives, controller, or software
RAID 50 (RAID 5 + 0)
RAID 50 is RAID 0 striping (no parity) across two or more RAID 5 arrays. The RAID 5 arrays provide the parity and data redundancy that level 0 is missing. You could conceivably leverage 5E or 5EE arrays as well. RAID 50 is faster than RAID 5 but requires six disks minimum. As with RAID 5, usable storage is 66% percent of total capacity.
- Pro: Increased sustained throughput, excellent safety
- Con: Long rebuild times, 66% usable capacity
- Requirements: 6 or more drives, controller, or software
RAID 60 (RAID 6 + 0)
RAID 60 is data striped RAID 0 fashion across two or more RAID 6 double parity sub-arrays. It offers twice the data redundancy of RAID 50 but, of course, requires more disks: four per RAID 6 array for a total of eight minimum. It’s faster than a single RAID 6. How much so depends on your pile and the number of RAID 6 sub-arrays. Usable capacity is 50% of total capacity.
- Pro: Increased sustained throughput, best data safety
- Con: Long rebuild times, 50% usable capacity
- Requirements: 8 or more drives, controller, or software
RAID 100 (RAID 10+0)
This is double-stacking RAID 0, or striping in RAID 0 fashion with no parity across multiple RAID 10 arrays, which themselves stripe data RAID 0 fashion across multiple RAID 1 mirrored arrays. Whew.
Obviously, with so many pipes, RAID 100 can be very fast while still maintaining the redundancy of mirroring. It requires eight drives minimum.
- Pro: Increased sustained throughput, best data safety
- Con: Long rebuild times, 50% usable capacity
- Requirements: 8 or more drives, controller, or software
RAID Level Attribute Comparison Table
Though we’ve spelled out the major differences in RAID levels, it’s always nice to see the basics compared in tight quarters. We’ve left out the non-starters.
Unlock cost-effective storage solutions with a RAID calculator.
