English Version

Raid 0, Raid 1, Raid 5, Raid 6, Raid 10 and other types of arrays

Artem Makarov aka Robin
Description of RAID arrays structure and methods of their repair

At the moment hard drives are the main data storages but unfortunately they do not show reliability we would like to see. One of the solutions of this problem is arrangement of RAID arrays from two or more drives. RAID arrays are of various configurations and their creation pursues different objects: from making a backup copy to acceleration of disk system.

Why data can disappear from RAID array?

The main reason that I face as a data recovery engineer is breakage of one or several disks of an array. The next in frequency problem with out-of-order RAIDs is breakage of RAID controller. Then, all kinds of controller snarks come, for example disk in an array becomes inactive, gets a “degraded” status, and logical failures, for instance logical volume or array configuration is lost. Often I have to face a human factor, e.g. array disks were interchanged, incorrect re-initialization of an array took place or wrong RAID rebuild was carried out.

In particularly complicated cases I have to deal with situations where RAID array was incorrectly assembled and after such a rebuild Windows checking utilities as check disk or others were started.

Let’s examine general types of RAID arrays:

RAID 0 RAID 0 description

RAID 0 or RAID stripe in elementary case consists of two disks, blocks of which alternate as follows: the first 64 kilobytes are on the first disk from 0th sector, the second 64 kilobytes block is on the second disk from 0th sector, the third block is again on the first disk right after the first one, the forth is on the second disk after the second block and so on. Block dimension may vary. In comparison with single disk such an organization of an array ensures capacity growth, as a result an overall performance of a disk subsystem increases.

RAID 0 description

RAID 0 description

When a failure of one hard drive happens, RAID array stops functioning so we should recover information from it.

RAID 0 redemption

How do we recover data from RAID 0? It is very simple. We should identify sequence of disks, dimension of bock, and then we provide virtual assemblage of RAID 0 with help of special software specifying beforehand all characteristics of the RAID array. At the end of this procedure it is possible to copy the information to external storage.

RAID 1 RAID 1 description

RAID 1. or RAID-mirror, mirrored RAID. The name contains its essence. All array disks have mirror copy of RAID content. Such a RAID array has higher fault-tolerance, it is able to work as long as at least one RAID array disks continues to operate.

Raid 1 description

Raid 1 description

RAID 1 repair

In spite of apparent simplicity of data recovery from RAID 1,  as it seems to be enough to recover the information on any storage, in fact engineer faces necessity of data recovery from the most urgent disk in the array, so as it is hard to determine which disk in whole array has had a fault first and accordingly contains older data; and which one has broken later and has more up-to-date information.

In the worst case I have to arrange an access to user information at all drives for data recovery from failed RAID array.

RAID 1e RAID 1e description

RAID 1e has a function of data blocks write-through (stripe). It means that every next block is written on following hard drive; moreover data block is duplicated on it from previous disk. Such system allows using odd number of disks in array. When one of the drives fails, no data loss occurs. Repair of RAID 1e is necessary if more than one disk fails.

RAID 5 RAID 5 description

In fact, RAID 5 is the same stripe but complete with blocks of checksum. Minimum quantity of HDD used for 5th RAID array arrangement is three. RAID 5 can be of following types: forward, backward, forward dynamic and backward dynamic. The difference between these types of RAID 5 is in checksum blocks sequence and their rotation. We should also mention of peculiar RAID 5 data recovery from HP servers (RAID 5 recovery Hewlett Packard); here controller organizes so-called delay after which rotation of blocks actually begins.

RAID 5 backward description

RAID 5 backward description

RAID 5 forward description

RAID 5 forward description

RAID 5 forward dynamic description

RAID 5 forward dynamic description

RAID 5 backward dynamic description

RAID 5 backward dynamic description


RAID 5 is able to function when one of drives in the array failed. In this case whole system slows down. Especially appreciable when working with database delays occur. If two or more disks are out of order, RAID 5 stops working; data recovery is necessary.

RAID 5 redemption

To recover data from RAID 5 array we need to clone, if possible, all drives of the array and then assemble them virtually. Procedure is the same as in case of data recovery on RAID 0: identification of disks sequence and dimensions of blocks, then on final stage of data recovery from RAID 5 detection of urgency takes place in those cases when one disk in the array had failed first, server was running in critical mode and only after that two or more drives failed.

RAID 5e RAID 5e description

RAID 5e array (RAID 5 enhanced) is an advanced version of RAID 5, it has higher performance and data integrity. In addition to space redundancy for checksum RAID 5e makes a reserve for hot sparing. Moreover, information is recording on every drive and redundant space is in each HDD as well. Thereby, speed of RAID array increases and every hard drive is used uniformly. To build RAID 5e at least 4 hard drives are necessary. A distinctive feature of this level array is that spare area is located in logical end of physical disk.

RAID 5-e backward description

RAID 5-e backward description

RAID 5e recovery

In comparison with RAID 5 5e type is assembling somewhat more difficult, as we have to take in account presence of hot-spare space. If the array did not lose disks and process of array self-recovery by means of controller did not begin, assemblage of RAID 5e array is not different from one of RAID 5. But in case hot-spare was used (or was begun using) we have to consider this nuance. Therefore, the basic approach is the same: identification of storages sequence, detection of block dimensions, its rotation type, parity and urgency of hard drives in the array.

RAID 5ee RAID 5ee description

RAID 5ee differs from RAID 5e only with its’ logical structure of data location. While spare area is allocated at the end of the array in RAID 5e, in RAID 5ee this area is divided to blocks which are rotated with blocks of checksum and data blocks. Such system of blocks arrangement quickens process of operability recovery when rebuild of the array taking place. RAID 5ee is also quite similar in its structure to RAID 6, here hot-spare block is used instead of one of the parity blocks.

RAID 5ee backward description

RAID 5ee backward description

RAID 5ee redemption

Peculiarities of data recovery from such a rare array as RAID 5ee consist in difficulty of its identification. In case the array failed because of RAID controller, identification is quite simple, presence of regularly repeating empty blocks of expected dimension speaks for it-self. If one hard drive failed in process and it was replaced by another one, then array rebuild process was initiated which was finished with an error; or RAID controller put the array into emergency mode, so previously empty blocks now may contain data calculated by controller on basis of checksum.

RAID 6 RAID 6 description

Subsequent logical development of RAID 5 is RAID 6, which is different from the 5th with two blocks of checksum; this means it is able to survive failure of two disks in the array. Rotation types of blocks are the same: forward, backward and their dynamic variations. When three or more drives fail in RAID 6 array, server stops operating – data recovery is needed.

RAID 6 backward description

RAID 6 backward description

RAID 6 forward description

RAID 6 forward description

RAID 6 redemption

Methods of data recovery are similar to ones of RAID 5. The problem is complicated with the fact that few developers of commercial software can boast with support of RAID 6 data recovery. But in general the procedure is the same: identification of sequence, blocks dimension and clarification of urgency degree. In addition, we should mention that slices, dozens of virtual machines, rare *NIX file systems and other things can be faced in RAID 6 arrays more often than in RAID 5, and even more often than in stripes or RAID-mirrors.

We advise to read additional material RAID 6 recovery.

RAID 10 RAID 10 description

RAID 10 is a combination of RAIDs of 0-level and 1-level. RAID 10 uses 4 (or more) hard disks, which are mirrored to each other by two (RAID 1) and pairs are united in RAID 0. Disk failure within the pair does not cause data loss, however when the pair fails information disappears, so data recovery is necessary.

RAID 10 description

RAID 10 description

RAID 10 recovery

Data recovery from RAID 10 array failed because of hardware failure (several hard drives in the array failure) is quite rare thing. More often they bring RAID 10 arrays with data lost because of controller failure or unskilled actions of staff (system administrators). Due to lack of effective utilization of disk space similar RAID arrays are used in big commercial organizations; and as a rule such tasks are complicated with slicers, it is staged presentation: physical hard drives – logical – physical – logical. When the hardware assembled virgin RAID array is divided into slices, which again are presented as physical drives, of which in their turn the other RAID array is assembled, possibly with a different configuration such as RAID 5 or RAID 0.


JBOD (Just a bundle of disks) is a tie of hard drives, strictly speaking, it is not a RAID array. Though some disks can be united in one logical section with help of OS or with hardware using RAID controller which supports the function of JBOD array build. Usually files are recorded sequentially to the end of the disk, then writing goes on to next HHD. Such data organization does not require any special equipment, as it was said above, and it can be implemented by means of OS or third-party software. Although, when one of the drives fails system is breaking down and data recovery from JBOD array in necessary.

JBOD array description

JBOD array description

JBOD Recovery

In case of JBOD disk array failure we should detect sequence of drives. This is not hard if JBOD array has just two storages, but task becomes more complicated if the array has three or more drives. In this case it is necessary to analyze file allocation table to determine the addressing of the file or directory, and then performing necessary calculations we can quite easily prioritize disks in the array. In addition, we should mention that in some cases when one or several drives of JBOD array failed and could not be recovered, the information from damaged JBOD was still available, though partially.

RAID 50 RAID 50 description

RAID 50 is a combination of two RAID arrays of 5th level united in stripe or RAID 0. Minimum quantity of drives for RAID 50 building is six pieces.

RAID 50 description

RAID 50 description

Picture shows a particular case of arrangement of such an array. It is necessary to take into account that some controllers have cascaded representation of disks in the chain physical – logical - physical. That is six physical disks are combined into two logical RAID 5 arrays (disks), then they are presented as two physical drives of the correspondent dimension and these drives are combined together in a stripe with its block dimension and sequence. The obtained logical drive assembled on technology RAID 50 is detected by OS means as physical disk and it is marked and formatted. As an illustration of possible accumulations we can imagine that this “physical” RAID 50 disk is divided into two logical ones by OS and of which then JBOD or RAID 0 are built. Engineer performing RAID data recovery and received six drives that were in such an array can spend a lot of time to build lookup tables of blocks and disks.

RAID 50 Recovery

As in case of RAID arrays of 6th level superstructure in the form of slices, dozens of virtual machines, *NIX file systems, VMFS and other joys of life are very common in RAID 50. Data recovery from RAID 50 is quite complicated task. To get started, we should try to get the most complete information about the intended configuration, the number and size of partitions, etc. Then, having defined the configuration, it is advisable to follow the assembly path which RAID controller has operated. That is RAID 5 arrays included in RAID 50 are assembled first, then they are uploaded into separate patterns which are combined into virtual stripe.

Which RAID array is better to use in practice?

Recommendations for RAID array choice and RAID array operating

To answer the question which RAID array is better to use in practice, we should determine task that office or system administrator face. Usually these tasks are two: increasing of disk subsystem speed and reliability of data storage. Occasionally one of these two objectives is dominated, as a rule, a combined approach with a shift of priorities towards performance or data integrity is used. An important role in making decision plays financial component which can be roughly represented as a price per gigabyte of disk space. It is obvious that stripe and JBOD has the lowest one as there is no loss, disk space of all HDD in the array is summed; while mirror RAID has the highest price for available gigabyte.

The most popular combined solution is RAID 5 arrays. These arrays are absolute leaders for use, respectively we face problems with RAID 5 arrays more often.

It reasonable to keep the most important data on RAID 1 or RAID 10 or RAID 5 with obligatory and wise politics of backup copies creation to external disks which are not a part of the array. Stripes are rational to use for allocated disk space in swap files on graphic stations, it is known that Photoshop, as an example, swaps back and forth, when handling large images can seriously affect system performance.

JBOD is typically used to reach the volume of a single disk that is not available as a physical one. That is if you want a disk of 10 terabytes for uploading a part of the image on 24 two-terabyte drives (and this is not the most complicated array we have to deal with), creation of JBOD array is the most convenient and correct approach to the problem in this situation.

RAID data recovery

If you face data loss from RAID array you can get a free online or telephone consultation and advice for further actions from data recovery specialist.

Write a comment
08 september 2012, 02:17

Hiso it did-not occur to you to buy a third and forth drives for when one drive filaed then. raid systems usually use a hot swap able drives for redundancy. the case of a little knowledge being a dangerous thing didn't spring to mind then. Copying the contents of those drives should have been paramount to a systems manager. if employed by a company dismissal would be immediate. so sadly you have lost something it could just be the power supply if it is you are lucky. but if both have filaed, what can one say .? there is another old saying in the computer world and it is this ITS NOT A CASE IF IT WILL FAIL BUT WHEN.