Understanding Basic RAID Definitions for
Commonly Used Types/Levels of RAID
Author: S.E. Fowler / Steve Fowler
Skip the intro . . .
It's outside the scope of our role in life as providers of data recovery service
to give full, complete definitions of all the elements crucial to the full understanding of RAID
data storage subsystems; researchers on this topic can find copious amounts of scholarly information elsewhere on the Internet.
It does seem incumbent upon us however, to give a deeper explanation, beyond the basic logical diagrams we provide
, for those certain features of common RAID implementations having impact upon the safety of your data
, as well as some of the salient terms
, if only for a basis for better clarity of communication among IT service personnel, supervisors, and other users.
Think of a RAID-"x
" as just one physical drive, because that is exactly what it looks like to the OS
, so yes, just as it's possible to have multiple hard drives on one system, it is possible to have multiple RAIDs on one system, or any combination of them. When data is recorded on ("written to") a RAID subsystem
, the RAID controller
slices it up into short blocks of data and places the uniform chunks consecutively on each individual hard drive
in the array
, resulting in what are called "stripes
" of data. In general therefore, a unique part of every data file is written to each member. This technique is referred to as "striping
", it actually results in a certain amount of I/O performance increase, especially during read operations, and it characterizes all the file data on every HDD within an array (except in the case of RAID-1). Striping is also essential to providing the basis for the fabled fault tolerance
of RAID (except RAID-0).
Explanations below cover all the commonly used types of RAID, from ZERO on up, except for the antiquated relics RAID-2, -3, and -4 which are theoretical variants of RAID-5, are rarely if ever used, and except for RAID-3, were never used to any significant degree even since the dawn of RAID's inception.
And our story is not complete without at least passing mention that there are actually quite a few additional, fairly esoteric types of RAID
covering virtually every permutation the human mind can conjure. Some are trademarked product entities of data storage equipment vendors, and others may be little more than intellectual curiosities, but neither will be further covered here. That's because by understanding just three elemental types of RAID you'll acquire a solid knowledge of technology inherent in all RAID subsystems, so please read on.
Three Elemental Types of RAID Defined
RAID-0 storage subsystems — striped to boost data transfer capability, especially in the form of external, highly portable Firewire (IEEE-1394) data storage appliances, have become extremely popular in recent times. Nonetheless, calling them a RAID at all is actually a misnomer because the "R" stands for "redundant", and there isn't any redundancy provided. It has been said that these configurations should be against the law! As we noted, they have no redundancy, they offer zero fault tolerance protection, and they are more likely to fail than one single "stand-alone" hard drive. Because they consist not only of high speed electronics, the constituent hard drives are also mechanical devices with high-speed, high-precision moving parts. This scenario makes the hard disk drive already the most failure prone component of everyone's computer. Now understand this: if a RAID-0 subsystem is made up of 3 drives, it is 3 times more likely to fail than any single drive; make a RAID-0 with 6 drives and it is 6 times more likely to fail than any single drive. The math is not hard.
RAID-0 has two or more independent drives. Originally RAID-0 was used to increase data I/O transfer speed, and while today some increase in performance may be realized, today's drives are so fast that the technique is just as likely to be a method used by product marketers to present the doubled or quadrupled storage capacity of two or four drives combined into a single box that connects to the computer system with the simplicity of one single drive — one really large capacity drive with one really large propensity to failure.
Two identical RAID-0's can be mirrored (see RAID-1 below) and the configuration will then acquire redundancy and be known as a RAID-10. RAID-0 is precisely equivalent to a RAID-5 after the latter has degraded as a result of a fault on one of its hard drive elements. The user of any RAID-0 configuration should live a computer user's life of greatly heightened data protection concern. For example, if your data protection scheme requires backup of data at least once per day to achieve a baseline level of data protection, a user of a 4 drive RAID-0 will need to backup four times per day to achieve the very same level of data protection. Yes, the math is starting to get more complicated now, if only a bit.
There is a common misconception about RAID-0 that comes from a lack of understanding about the nature of data striping. For example let's say, keeping it simple, that we're considering a dual-drive RAID-0. The mistaken belief is that even though one of the two drives has catastrophically failed and cannot be recovered, at least the data on the other drive can still be used. Not so. To implement striping, every single file on the array has been sliced up into small segments by the array controller, and as each stripe consists of two segments (in this example), only half of each file is accounted for on the un-failed drive, and that half contains only alternate pieces or every other slice of the original data. Because of this fact a data recovery company will always require all of the array drives, working AND non-working, to effect recovery of usable data.
Moreover, if damage to the recording media has occurred on the failed drive in the same physical location as the most critical file stored upon the RAID-0 array (again keeping the example with just two drive elements), even though half of it is completely unscathed as it rests upon the good drive, in general no usable part of this file will be recoverable. Reasons pertaining to this reality are precisely what causes RAID-0 to occupy the statistically highest percentage of non-recoverable data storage subsystems.
It's not unwise to consider all hard disk drive data storage as temporary storage, because when a drive fails, sometimes the information it holds almost literally disappears in a instant, possibly forever. The more drives that make up an array, the more likely array failure becomes. In view of all the forgoing, if you are deploying a RAID-0 with eight hard drive elements, it is quite unwise NOT to consider the entire array contents as temporary storage -- VERY temporary. Please measure your exposure to data loss very carefully.
RAID-1 is also known as "mirroring"; the array will always have an even number of drives. RAID-1 is the simplest and easiest of all the levels to understand because typically, striping is not used. All data stored into the host computer system is written to each of two drives or subsystems automatically, thereby creating two copies of all data: an automatic, up-to-the-second backup of every bit, hence the analogy referring to a "mirror". This level of RAID is also arguably the most reliable means of data protection available, but the trade-offs may turn out to be substantial.
One quick point which actually applies to all levels of RAID. The purpose of redundancy is not to replicate data. It's purpose is to increase or protect the availability of data, and availability is impacted primarily by data storage device function, not by merely having more than one copy of a file. Therefore, the protection provided is always a matter of preventing data loss in the event of breakdown or failure of a hard drive (or whatever storage device). It's patently obvious once it's pointed out, but if two logical drives were to be created (i.e., by setting up partitions) on one single, physical hard drive and one was a mirror of the other, even though there are two copies of all data, when that one hard drive fails, access to all those copies is simultaneously lost.
RAID-1 takes hits in areas of both cost effectivity and perhaps performance. Depending on design of the RAID controller used, read operations can provide higher performance by reading from both elements. Since these "hits" are of concern not so much to an individual, but more likely only to IT administrators of large commercial enterprise data archives, with today's low cost of hard drives RAID-1 is arguably the best of all possible data protection schemes for the small-business or home computer user.
The most commonly used, prevalent and widely implemented redundant disk drive storage subsystem for commercial data processing and business continuity (BC) is the RAID-5. RAID-5 configurations must have three or more individual or independent drives, a.k.a. members, and can tolerate even total failure of any drive in the array without losing a single byte of stored data. In a "write" operation, data to be recorded is striped across all array members interspersed with a parity block and distributed so as to place one segment containing the parity checksum on a different drive within each striping cycle. Sometimes defined as rotating parity, this scheme is the reason why it doesn't matter which drive in the array fails. The RAID-5 subsystem often provides the best compromise between performance, fault tolerance, and cost efficiency.
While striping makes read operations faster, calculating and adding the parity block to each stripe can slow write operations by as much as 50%. The upside for this invention is support for read data operations competitive with RAID-0, in addition to allowing for complete failure of one of any hard disk drive element, while the subsystem itself continues to operate without losing any data or operational uptime. The name for the capability provided by this characteristic feature of RAID is called fault-tolerance.
If any one of the hard drive array member elements fails, the RAID-5 subsystem controller detects it and a signal is sent to the computer system it's hosted by and makes the fault known. Precisely at the moment of failure, the array's status becomes degraded and the subsystem's inherent fault-tolerance is lost as its one and only layer of protection evaporates. Despite the drive failure however, all of the stored information remains fully available. The failed drive can then be replaced, and once the replacement drive has been fully reintegrated into the array by means of a successful rebuild operation, fault-tolerance is reinstated and the maximum data availability supported by the subsystem design is once again restored.
Before the faulted drive is replaced the subsystem operates as though it is a RAID-0. In this degraded mode, fault tolerance is not in effect and that is cause for concern about the loss protection factor regarding your data. Without a backup, your need for data recovery service becomes imperative in the event of a second HDD failure. When a "hot spare" device has been implemented within the RAID-5, diagnosis and rebuilding of the array for fault tolerance reinstatement can be automatic, and generally does not require downtime of the data storage subsystem. It's worthwhile to note emphatically, if a second drive failure occurs prior to successful rebuild, data access is lost. If more than one drive failure has occurred, rebuilding is not possible and data access has been lost. In either case data recovery expertise will be required.
Even though recovering the data from a failed RAID-5 is considerably more challenging than recovery of a stand-alone desktop hard drive (which is challenging in itself), because of the inherent redundancy of data being stored, the actual recoverability is in point of fact better, as long as the subsystem is presented in its "as failed" condition. Information is provided elsewhere on this website to help RAID-5 users establish procedures and launch initiatives promoting the protection or availability of stored data. [CLICK HERE] Ill-advised remedial actions often result in overwrites of critical array information, loss of array structure, and may preclude data recovery. This means permanent data loss.
Beyond Elemental RAID
— Extended Redundancy
Much less well-known than RAID-5, RAID-6 can tolerate the concurrent failure of no less than two hard drives while precluding data loss and system downtime. This feat of increased availability is accomplished by applying a second layer of redundancy by means of two separate, independent parity blocks within each stripe written to the data storage subsystem array, distributed among each of the active members in the array. Thus, another moniker for RAID-6 is "double-parity RAID".
This implementation requires a minimum of four individual drives to form an array, but will typically have a significantly larger number of members since performance, fault tolerance, and cost efficiency are all improved relatively with a larger population. With a solid understanding of RAID-5 theory, the conceptual structure of RAID-6 can be easily grasped by simply imagining the addition of one more hard drive to such an array. This drive is used to provide space needed for the additional parity information, so that the available data storage capacity between the compared arrays of this example remains unchanged. In actual practice of course, a RAID-6 controller is also required.
Many of our data recovery clients relate tales of lament, saying how a second HDD failed prior to completion of a RAID-5 rebuild activated in response to the RAID controller notification that the array has degraded as a result of an initial HDD element failure. The second layer of redundancy provided by RAID-6 offers protection against this scenario.
There are but two trade-off's with RAID level six. The first is an I/O performance hit due to array controller processing to calculate the second parity check-sum, along with drive response time to record the second set of parity information. This action typically decreases data write performance about 20% below the (already reduced) performance of RAID-5. As with RAID-5, the read-back performance is fast because the parity information is not part of the file data.
The second, not so important trade-off is a hit in storage efficiency because of the one additional hard drive needed to supply the capacity for containing the second parity information set. In practice, this is hardly any concern at all, for given the cost of drives today it's a small price to pay on this insurance policy for added protection. Because the likelihood of a drive fault increases in parallel with the number of member drives used in the array, this configuration really starts becoming a compellingly attractive solution for increasing data availability when array member populations above six or seven drives are used.
The RAID-TEN is two ideally identical arrays; the setup consists in the mirroring (RAID-1) of two RAID-0's; alternatively this same configuration may be referred to as "RAID 1 + 0" (or "0 + 1"). Since it is a combination of a RAID-1 and a RAID-0, it seemed natural enough to drop the arithmetic operator and simply use the number ten. A minimum of four hard drive elements is required, and the total number of array members will always be an even number.
The RAID-10 configuration has a practical fault tolerance no better than any RAID-5, may have higher maintenance costs, and most certainly has lower efficiency of storage device use. However, by means of data striping, subsystem performance should be better than RAID-5 because there is no overhead for parity calculation.
Because this configuration does involve RAID-0 use, implementers should be aware of all the various drawbacks inherent in all zero-redundancy storage device array schemes (see RAID-0 warnings above). While this type of array continues to provide redundancy with one or even more failed drives — within the same RAID-0 subset, if more than one drive becomes concurrently failed on both sides of the mirror, all the data on the RAID-10 subsystem becomes unavailable, and will typically require data recovery service to restore access.
Author: S.E. Fowler / Steve Fowler
Call Now for RAID Recovery Service you can rely on . . .