Customer Support: 1.866.933.9800

Sales Questions: 1.866.469.6120

What RAID Set should I choose for my Dedicated Server?

So, what RAID set should you select for your dedicated server? This is another question we commonly encounter, especially for those clients who are moving to their own dedicated server (or servers) for the first time. We also see this question asked by new clients after they have suffered a critical data loss from a hardware failure. Before deciding what RAID set to use on your dedicated server it is important to understand what RAID is and how it operates.

Let's first address, "What is RAID and what does RAID stand for?" Today RAID is an acronym that stands for "Redundant Array of Independent Disks." Historically, however, RAID was originally used as an acronym for "Redundant Array of Inexpensive Disks." In essence RAID is a storage technology that links or combines multiple hard drives so that data can be stored on them as if they were one logical unit. So, "What is RAID?" - in its simplest form, RAID takes multiple physical disks and makes them appear and function as one single drive.

Now that we understand what RAID is, let's take a close look at how it operates (RAID sets or levels). What is a RAID set? RAID sets are different schemes or architectures that divide and replicate your data amongst the various member physical drives that are present in your RAID array. Each RAID set provides a uniquely different balance by addressing the storage characteristics of resiliency, performance, and capacity. There are a number of standard RAID sets that have emerged, ranging from RAID 0 to RAID 10.

1) What is RAID 0?

RAID 0 is block-level striping on 2 or more drives. This occurs without parity or mirroring; and subsequently, the RAID set offers no redundancy. Hence, if one of the member drives fail, all of the data on all of the drives is lost. This is the reason why you would likely never want to run RAID 0. For each additional drive you add to your RAID set you are mathematically increasing your probabilities of catastrophic data loss.

So the next logical question - why use RAID 0? The answer is found in the performance gained without the loss of drive capacity. Because the data is cut and striped on 2 (or more) drives, the data can be accessed or written in parallel - approximately twice as fast. This is measured in IOPS which is the standard performance based unit of measurement for Input/Output operations per second. So why do people choose RAID 0? The answer is quite simply that they seek the speed and performance gain of striping (higher IOPS) and are willing to tolerate the data loss in the event of a drive failure. A common use of RAID 0 would be a Photoshop image retouching station.

2) What is RAID 1?

RAID 1 is data that is written to 2 (or more) drives that is mirrored without parity or striping. So why use RAID 1? The primary reason is the member drives in a RAID 1 array essentially produce a mirror of each other. Hence, if 1 of the member hard drive fails, no data loss is suffered as the other member is a mirror of the lost drive. What is the downside of RAID 1? Simply, no write performance is gained by the use of 2 member hard drives. Given the advances in operating system efficiency, read performance is usually increased as the operating system will direct simultaneous reads of both hard drives to retrieve the data quicker. In the case of a 2 drive array, this is commonly known as "duplexing". In RAID arrays with more than 2 member drives it is referred to as "multiplexing". Why choose RAID 1? Simple (literally!) - RAID 1 is a simple technology that is a popular choice on dedicated servers because of the additional protection from data loss along while affording some increased read performance.

3) What is RAID 2?

RAID 2 is bit-level striping with dedicated hamming code parity (hamming code parity is a form of error correcting code or ECC). Doesn't make sense, don't worry, RAID 2 is rarely used because it does not use the standard techniques of mirroring and striping. RAID 2 splits data at the bit level then spreads it over a number of member disks. Disk spindle rotation is synchronized and the data is striped so each sequential byte is on a different drive. RAID 2 requires a minimum of 3 member drives - one drive may be lost without suffering data loss. RAID 2 provides large gains in read performance because of the large number of member drives (up to 39!) that may be accessed in parallel. Write performance is poor due to the bit level striping and the ECC calculation that is taking place. Why is RAID 2 never used? RAID 2 is rarely used as the controllers are rare, complex, and expensive.

4) What RAID 3?

RAID 3 is byte level striping (data blocks are subdivided and striped on 2 or more drives) with dedicated parity using a dedicated parity drive. RAID 3 requires a minimum of 3 member drives but more typically uses 4. Given that parity is used, RAID 3 has a fault tolerance of 1 drive. Why choose RAID 3? RAID 3 provides large read and write performance gains, especially for large file block transfers. It is for this reason RAID 3 is often used in Video streaming, publishing and editing. What are the disadvantages of RAID 3? Unfortunately, RAID 3 is rather inefficient with smaller files and I/O operations. Raid 3 technology is also considered complex and expensive; therefore it is not too common.

5) What is RAID 4?

RAID 4 is block-level striping with dedicated parity. RAID 4 is exactly the same as RAID 5 except parity is dedicated and limited to one parity disk. RAID 4 is similar to RAID 3 except that striping is done on a block-level basis rather than a byte level. Why use RAID 4? Read performance is accelerated given that I/O requests are performed in parallel. In addition, the size of the stripes for striping can be adjusted depending on the application being served. This results in very fast random and sequential reads. As well, RAID 4 does not require the spindles of disk to be synchronized as in RAID 3, making the technology less complex and more economical. As in other RAID sets a single member drive can be lost without the loss of data. Why not use RAID 4? RAID 4 is not commonly used because of the poor performance of random writing. The poor performance is a result of parity calculations that lead to a bottleneck created by the single parity disk. Today, RAID 4 is not commonly used in many dedicated servers.

6) What is RAID 5?

RAID 5 is block-level striping with distributed parity. Striping and data are distributed by independent read and write operations (not in parallel like RAID 3). Instead of using a dedicated parity disk (such as RAID 3 and 4) parity is distributed across all the drives and a minimum of 3 disks are required for a RAID 5 array. Why choose RAID 5? Quite simply RAID 5 is the most secure RAID level that offers accelerated read performance. Further, in RAID 5, a single disk can be lost without encountering any data loss. RAID 5 is considered to be one of the best all-around systems that combine a high level of security, accelerated performance, and an extremely efficient use of storage. In many instances RAID 5 is a perfect combination and it is for these reasons we recommend the use of RAID 5 on many of our dedicated servers. RAID 5 has become one of the most popular levels as it is simply the ideal RAID solution for many server applications.

7) What is RAID 6?

RAID 6 is block-level striping with double distributed parity amongst its member drives. RAID 6 is similar to RAID 5 except that RAID 6 creates two parity blocks for each data block. Therefore, a RAID 6 system is not as vulnerable to data loss as it can suffer the hardware fault of 2 of the member drives. RAID 6 requires a minimum of 4 member drives. Read performance of a RAID 6 array is accelerated; however, write performance is relatively poor given that for each block of data 2 parity blocks must be written. Why use RAID 6? RAID 6 is best suited for data where a higher degree of reliability or certainty is required. An example of this would include critical databases, applications, E-mail and web servers. So the obvious question, why not use RAID 6? The answer: RAID 6 is expensive. RAID 6 requires a high availability system (for drive members) and a very sophisticated controller - that alone may be cost inhibitive for many. Further, given the dual parity, the demand for disk space is large.

8) What is RAID 10?

RAID 10 is a combination of RAID 0 and RAID 1 into one single system and is commonly thought of as the "striping of mirrors". RAID 10 affords security by mirroring all data on a secondary set of disks while striping across each set of disks to increase read and write performance. RAID 10 requires a minimum of 4 member disks. The addition of further members must be paired to account for the mirroring that will take place. Why use RAID 10? RAID 10 offers increased write operation performance and data redundancy is more robust than RAID 5 (multiple drive failures are tolerated). Why not use RAID 10? The read performance of RAID 5 is faster and the storage capacity is less efficiently used as the data is mirrored. Overall, if the higher redundant cost overheads of RAID 10 can be tolerated, RAID 10 offers very high performance with some of the highest level of data protection possible.

So, what RAID level is the best? A better question is, "What RAID level is best for you?" As you can see this question is not an easy one to answer. Ultimately, your RAID level selection will depend on your budget (hardware availability) and how to meet your key goals of performance, resiliency and storage capacity.