RAID 0

RAID 0 is simple data striping. Rather than storing data all in one chunk on one drive, striping deposits small pieces of the data across two or more disks. RAID provides high performance at relatively low cost, but doesn't actually provide any data protection. If the system loses even one disk, it loses all data, because a chunk will be missing from every file stored.

RAID 0 is used in applications where peak transfer rates matter and data protection is less important, according to EMC's Burke. Video editors and small engineering groups may use RAID 0 alone. "It works when the master version is somewhere else and can be easily retrieved."

Though occasionally used in a stand-alone environment, RAID 0 is most often used in conjunction with other RAID levels, especially RAID 1 (mirroring). This combination is sometimes known as RAID 10 and sometimes as RAID 0 + 1, depending on what company is doing the marketing.

RAID 1

RAID Level 1 is simple data mirroring--it duplicates all data from one drive to a second drive, providing 100 percent redundancy: Should one drive fail, the mirror drive contains the exact same data. It provides optimal data protection but punishes the user by doubling the cost of storage. It's among the most expensive systems per megabyte since all data has to be written twice.

RAID Level 1 is used where price is of secondary, or even tertiary, consideration. Banks and other financial institutions--whose data literally is money--need the exceptionally high performance and reliability RAID Level 1 can provide. As the price of storage continues to fall, RAID 1 will probably continue to grow in popularity.

RAID 2

RAID Level 2 uses a different method of striping data, incorporating a dedicated checking disk and including error correction code across disks in the array. It provides a fair degree of data redundancy through the use of this "Hamming code." It's pretty much an obsolete technology though, more suited for large data transfers than transaction processing.

"You can pretty much forget Level 2," said Aberdeen's Hill. You'd be hard pressed to find a user today--the only reason to even mention it is that it's been assigned a low RAID level number: If it wasn't mentioned, people would keep asking "what happened to Level 2?"

RAID 3

RAID 3 stripes data across three or more disks. It differs from RAID 1 in that it features an extra disk dedicated only to parity bits. The parity disk provides an extra level of security not found in RAID 1: if a disk fails, the data it contains can be reconstructed from the data on the parity disk.

RAID 3 provides high bandwidth for moving large blocks of data quickly, but isn't as efficient in moving smaller, random blocks of data. As a result, it's used primarily in certain specialized applications, such as video production, as well as scientific applications including seismology and genetics.

RAID 3 systems represent the middle ground: performance and reliability are pretty good. They do lose 25 percent performance to overhead, since one of every four disks isn't used for storage. It's not as reliable as a completely mirrored system as found in RAID 1.

RAID 4

Like Level 2, RAID 4 mostly takes up space in the RAID lineup. "It's almost never implemented these days," said EMC's Burke. Similar to RAID 3, RAID 4 stripes data in blocks across three or more disks, plus one parity drive. Data is striped by blocks.

RAID 4 has a very high read rate for blocks of data, and a terrible write rate, according to some experts. It's also difficult and inefficient when it comes to data recovery. Disk overhead is relatively small, since only one disk is used for parity checking.

Users who query a database of large files regularly might take a look at RAID 4, but they'd probably be better served by RAID 3 or RAID 5.

RAID 5

RAID Level 5 stripes data across three or more drives like RAID 1. But it also distributes parity data. Parity checking adds an extra, ninth bit to a byte of data based on the content of that byte. The parity information can be stored elsewhere; then, if part of the original byte is corrupted, the parity data can be used to reproduce the byte with which it's associated. The system also has an extra drive where the parity data is duplicated.

RAID 5 is probably the most popular of the RAID levels, with the possible exception of Level 1. "RAID 5 is now standard," said Aberdeen's Hill. It provides a high level of data protection, is reasonably good at data recovery, and has relatively low overhead in terms of disk capacity. It's usable for most applications, especially those such as databases with small, random data requests; Web servers, for example, are prime candidates for RAID 5, since virtually all their work is read transactions, rather than write.

RAID 10

RAID Level 10 (and its counterpart, RAID 0 + 1) combines the strengths of the first two RAID levels. "It's literally striping added to mirroring," explained EMC's Burke.

In a typical four-disk system, two disks are established as mirrors of one another (RAID 1) and then that data is striped across two additional disks (RAID 0). The result is a system that is reliable enough to survive concurrent disk crashes, and faster than most storage systems because I/0 requests don't have to go to a single drive. It's comparatively costly, though.

RAID 10 is useful for systems where money is no object or, at the very least, budgetary considerations are secondary to speed and performance. Small offices and some database servers may look to RAID 10 as a possible storage solution.

RAID 0 + 1

Like RAID 10, RAID 0 + 1 combines the strengths of the first two levels of RAID. If RAID 10 is striping added to mirroring, "RAID 1 + 0 is mirroring added to striping," said EMC's Burke. Basically, data is striped across multiple disks (as in RAID 0) and each of these disks is mirrored (RAID 1).

In terms of reliability, RAID 0 + 1 is about equivalent to RAID 5. It's considerably faster than RAID 5, though, since I/0 requests don't have to go to a single drive. The system isn't easily scalable, however, and is generally recommended for applications such as imaging and general-purpose file serving.

There are some other exotic flavors of RAID--levels that are off the beaten track, and in some cases actually proprietary solutions. Here are a few of them:

• RAID 6: Similar to RAID 5, RAID 6 does a second, independent parity computation. The most reliable RAID solution available, it's expensive and slow.
• RAID 7: Like RAID 4, RAID 7 has a proprietary interface that controls the I/O. It's a proprietary solution of Storage Computer Corp. It delivers high performance but is very expensive compared to other RAID solutions.
• RAID 53: This should really be called RAID 03, because it's implemented as a striped array whose segments are RAID 3 arrays. It's much higher performance than RAID 3, but is quite expensive--as it requires two parity disks--and tricky to implement.
• RAID 100: "This level is similar to RAID 10," according to Digi-Data's Tomeo. It mirrors and stripes data, and then adds a second level of striping down the channel. It offers a 10 to 15 percent improvement in performance, he estimates, with a 20 percent increase in cost.

I “borrowed” this from storagepipeline.com.

I will give it back to them when I am done with it.