Let’s start with something small: “RAID-array” or popularly “RAID”, what is it?

RAID– this is an abbreviation that stands for yak (English: “Redundant Array of Independent Disks”), which in Russian translation is “redundant (reserve) array of independent disks.”
In simpler terms, a “RAID array” is a combination of physical HDD drives into one logical drive.
Logical disk– the primary HDD disk is divided into several logical ones. So, it’s easy to get stuck on stationary computers, and it’s a lot of money.
As it turns out, the primary physical disk can be divided into a number of logical disks. With "RAID" everything is reversed - a number of HDD drives are installed in a separate element (to save money), and then the operating system accepts all HDD drives as one. So, the operating system is 100% loaded, so there is only one physical disk connected to it.

There are two types of RAID arrays, hardware and software:

1) Hardware RAID arrays- Be sure to proceed before the operation of the operating system is carried out with the help of specialized utilities installed (hardwired) into the “RAID controller” - just like the “BIOS” symbol. After this procedure, when the “RAID array” is connected, the operating system at the installation stage will load your HDD drives as one.

2) Software RAID arrays- Created for additional connection of HDD drives to any operating system. Then, when you connect HDD drives, there is a number of physical drives and only with the help of the Operating System, with the help of software, the HDD drives are combined into one array. The OS itself is not installed on the RAID array itself, but is installed before the array is created.

"What is this for?"- You don’t have food! The answer is simple: in order to increase the speed of reading and writing data and increase security and visibility.
Let’s take a look at how a “RAID array” provides greater productivity and security for your data?” – To get the point across, let’s take a look at the different types of “RAID arrays”, how they are formed and what comes out of them.

Let's take a look at "RAID-0":

More than one HDD disk is required for additional daisy chain connection, after which obligations are required. - if you take a bunch of HDD disks, each with a volume of “500GB” and create “RAID-0” from them, then the operating system will easily install HDD disks as one of them, from which we take one HDD disk with a volume of 1000G b( 1Tb) . Once the number of disks is combined into one array, the read and write speed of the storage device will be twice as high, while the disks will have less speed.

butt- The database has been expanded on two physical HDD disks, one of which will work only to read data, while the other customer will write data to another HDD disk, and all will work at the same time. And since the data base will be distributed only on one disk, the HDD disk itself will have the function of reading and writing from different computers and will sequentially contain its own software. The "RAID-0" array gives the ability to select read and write in parallel. For speed, you can create as many as your RAID-0 array of HDD drives, multiply. N = number of installed HDD disks, array.

The RAID-0 array has only one shortcoming, this minus overlaps everything, and the advantages of this alternative are that in the RAID-0 array there is daily resistance up to wipes. The problem lies in the current situation, if one of the physical HDD disks installed in the array goes wrong, then the entire array dies.
There is an old joke from this drive: “What does “0” mean in the RAID-0 value? - about the information that is updated after the death of the array!” (though it’s not fun at all, since it’s even more important there).

Let's take a look at the "RAID-1" array:

A dozen or more HDD disks are combined into one by installation in a specialized array, then. How to take a bunch of HDD disks with a capacity of 500 GB and create a “RAID-1” array, The operating system is compatible with one 500GB array.
The read and write speed of the RAID-1 array will be exactly the same as in one HDD disk, since reading and writing will be carried out simultaneously on both HDD disks.
The “RAID-1” array does not increase the speed of vibration, but you are guaranteed resistance to damage if one of the HDD drives comes out in order, and there will be a further backup (backup copy) of information on the other HDD drive. Once the data from the array is completely straightened, the data is retrieved from both disks at the same time!

Let’s take a look at the “RAID-5” array:

The most secure option for RAID-5. Filling the array with information is done by expanding the formula “(N - 1) * DiskSize”, where N number is the number of HDD drives in the array, and the abbreviation “DiskSize” is the number of each installed HDD drive, then. When we create an array of the "RAID-5" version with 3 HDD disks, each with a capacity of 500GB, we have a memory array of 1000Gb 1 terabyte.

The essence of the RAID-5 array is the same - a number of HDD disks are combined into “RAID-0”, and on the third HDD disk (which is not covered by insurance) a so-called “control sum” is saved - this information is intended for updating one of the disks of the array y, At the time of death, the “RAID-5” array has a little less write speed, leaving only a small hour spent on unpacking and writing the extracted sum to the additional disk, and the read speed is lost as much as the “RAID-0” array.
If you use one of the HDD drives in a RAID-5 array that you have, the speed of reading and writing will sharply decrease, since all operations that are performed are controlled by additional data controls active actions.

In fact, RAID-5 is converted to RAID-0 and since there is no need to worry about updating the RAID array, there is a risk of wasting the data.
In parallel with the “RAID-5” array, you can vikorize the “Spare disk” - a spare one. During the stable operating mode of the “RAID array”, the “Spare” disk is not corrupted and is in idle mode. Should there ever be any critical situation, the backup “RAID array” will be updated automatically – the backup HDD disk will receive updated information from the damaged HDD with the help of checks and balances. zhennya of which appear on the fixed HDD disk.
The “RAID-5” array can be created from at least three HDD drives and will help you to restore your data only in one batch. When different defects appear on different HDD disks overnight, the RAID-5 array does not rotate.

Dali array "RAID-6":

There may be increased capacity when aligned with the “RAID-5” array. By the way, the essence of the work is the same, as with the “RAID-5” array, only the calculation of control sums is carried out not on one HDD disk, but on two HDD disks, and the entire subset of control sums is based on completely different algorithms, which harmonizes with the inherent shift in flexibility. everything " RAID array" zagalom. It is important to assemble a RAID-6 array from 4 HDD drives. The formula used to sort out the size of the memory array looks like this - (N - 2) * DiskSize, where N is the number of HDD disks installed in the array, and “DiskSize” is the size of the memory of each HDD disk, then. When you create a “RAID-6” array with five HDD disks with a nominal value of 500GB, you will end up with an array of 1500Gb (1.5Tb-terabytes).
The speed of the “RAID-6” array when recording will be lower than that of the “RAID-5” array by approximately 10-15%, the decrease in speed is due to additional expenses for the hour of decomposition from the recording of control sums.

Array "RAID-10":

These types are called “RAID 0+1” or “RAID 1+0”, which is a symbiosis of “RAID-0 and RAID-1”. This array is created from at least four HDD drives: on the first partition “RAID-0” and on the other “RAID-0” in order to improve the speed of reading and writing, which will mirror the array “RAID-1” "- tse required to improve durability up to 100% strength. The "RAID-10" array could combine the advantages of the first two options - with the best possible speed and flexibility.

The "RAID-50" array is an analogue of the "RAID-10", which is a symbiosis of "RAID-0 and RAID-5" - in fact, it is collected as a "RAID-5" array, only the storage elements that are included before will not be Physical HDD drives are stored in arrays in the “RAID-0” plan. Thus, the “RAID-50” array will give you the miraculous speed of reading from writing and the acceptable stability and reliability of “RAID-5”.

Dali array "RAID-60":

The same principle: in fact, it is not “RAID-6”, it is collected from many “RAID-0” arrays.
There are other combinations of arrays, such as “RAID 5+1 / RAID 6+1” - in essence, they are similar to “RAID-50 / RAID-60” with the difference that the base of their array elements is not “RAID- 0", like the others, and mirrored the "RAID-1" array.

Understanding about combined “RAID” arrays:

Essentially the same arrays as "RAID-10" / "RAID-50" / "RAID-60" and "RAID X+1"- these are direct reductions of base arrays such as “RAID-0” / RAID-1 / RAID-5 and RAID-6 – mainly they are used in order to increase either read speed, or write speed, or to increase durability up to idmov, vikorystyuchi contains the standard functionality of basic, standard types of RAID arrays.

If we look at it from a practical point of view and discuss the stagnation of any “RAID arrays” in life, then the logic behind everything is simple:

1) RAID-0 array in pure appearance there is no vikoryism (at all!);
2) "RAID-1" The array is mainly used where readability and write speed do not play a particularly important role, and the greater need is for resistance to damage - but for example: on the “RAID-1” array it is good to install various system operations and. In this case, the HDD disks themselves do not require any operation, the HDD disks themselves are sufficiently flexible for work, and resistance to damage is ensured;
3) RAID-5 Installed where you need fluidity and resistance to damage, but there is no possibility of adding a large number of HDD disks, or if there is a need to update the arrays, there will be damage when Therefore, so that the robot itself does not get stuck on the array itself, Spare disks will help in this situation.
4) Standard design for the RAID-5 array:
At the same time, this is what the NAS server is called;
5) "RAID-6" array:
There is a threat that only a few HDD drives in the array can go wrong. In practice, this practically does not exist, as in paranoids;
6) "RAID-10" array:
Stick there where fluidity is required, for fluid work it is safe. So the main purpose of the “RAID-10” array is database servers and file servers.

The axis is the principle and everything that I wanted to understand, what and why!

RAID(English) redundant array of independent disks - superfluous array of independent hard disks)- an array of numerous disks, coated with a controller, interconnected by sliding channels and compressed by the external system as a single unit. Depending on the type of vicorized massif, different stages of fluid capacity and speed can be provided. Serve to increase the reliability of saving data and/or increase the speed of reading/writing information. Initially, such arrays were used as a reserve for storage on random access memory (RAM), which was expensive at that time. Then, the abbreviation took on a different meaning - the array was already made from independent disks, due to the importance of choosing multiple disks, rather than sections of one disk, and also the road (now obviously just a few disks) possession necessary for wake up massif.

Let's take a look at what RAID arrays are like. Let us first look at the equals represented by those from Berkeley, then at their combinations and non-primary regimes. Please note that if discs of different sizes are used (which is not recommended), then pay attention to the work of the smallest. The greatest promise of great disks will be inaccessible.

RAID 0. Disk array from stripes without capacity/parity (Stripe)

The data is massively divided into blocks (the size of the block can be set when the array is created) and then written to the surrounding disks. In the simplest form, there are two disks, one block is written to the first disk, another to the other, then again to the first, and so on. This mode is also called “scratching”, because when recording blocks of data, the disks on which the recording is being made are scratched. Apparently, blocks are read the same way. In this way, a parallel input/output operation is carried out, which leads to greater productivity. Whereas previously we could acquire one block in an hour, now we can obtain several disks at once. The main advantage of this mode is the high speed of data transmission.

However, miracles don’t happen, and when they happen, they don’t happen often. Productivity does not grow by N times (N is the number of disks), but less. First of all, access to the disk increases N times an hour, and is so high compared to other subsystems of the computer. The controller's brilliance does not diminish. If it is not the same, the liquid may be slightly reduced from the liquid of one disc. This also includes the interface that is the RAID controller for connecting to another system. All this can result not only to a lesser, but also to an increase in the speed of linear reading, but also up to a number of disks, installing more for any increase will no longer be allowed at all. Or, for example, the liquidity will be reduced a little. In real situations, with a large number of drinks, the chance of encountering this phenomenon is minimal, because the fluid is already strongly pressed into the hard drive and its capacity.

Apparently, this regime has no such supernaturalism. The entire disk space is captured. However, if one of the disks goes wrong, then, obviously, all the information is wasted.

RAID 1. Mirror

The essence of this RAID mode is to create a copy (mirror) of the disk with the method of increasing the stability to RAM. If one disk goes out of order, then the robot does not stop, but continues with one disk. This mode requires a couple of disks. The idea of ​​this method is similar to backup, but everything is backed up “on the fly”, just like updating after a failure (which is sometimes even more important) and there is no need to spend an hour on it.

Disadvantages - high overhead, which means you need twice as many disks to create such an array. Another disadvantage is that every day there is any increase in productivity - even a copy of the data from the first one is simply written to another disk.

RAID 2 Array using the best Hamming code.

This code allows you to correct and detect additional damages. Actively competes with memory correction correction (ECC). In this mode, the disks are divided into two groups - one part is stored to save data and operates in the same way up to RAID 0, dividing data blocks across different disks; Another part is used to save ECC codes.

The advantages include on-the-fly correction of data transfers and high data streaming speed.

The main drawback is the high volume (with a small number of disks, the difference is greater, n-1). With an increased number of power disks, the number of disks saving ECC codes becomes less (the power overhead decreases). Another disadvantage is the low speed of working with fractional files. Due to the bulkiness and high overhead with a small number of disks, this RAID level is currently not being promoted, giving positions to higher levels.

RAID 3. Vidmovostoy array with bitumen and parity.

This mode writes block data to different disks, such as RAID 0, or adds one more disk for parity. Thus, the overhead is much lower than in RAID 2 and there is less than one disk. If one disk fails, the fluidity practically does not change.

The main disadvantages include the low speed when working with large files and the lack of typing. This is due to the fact that all control codes are saved on one disk and during input/output operations they must be rewritten. The fluidity of this disk combines the fluidity of the entire massif. These pairings are written as soon as the data is recorded. And by the time of reading, the stinks are checked. As a result, there is an imbalance in the read/write speed. At the same time, reading small files is also characterized by low speed, which makes parallel access from independent disks impossible if separate disks are being written in parallel.

RAID 4

Data is recorded in blocks of different disks, one disk is vicorized to preserve parity. The flexibility of RAID 3 lies in the fact that blocks are divided not by bits and bytes, but by sectors. The advantages lie in the high speed of transmission during the hour of work with large files. The productivity of the work is also high with a large number of requests for reading. From the shortcomings, it can be seen that RAID 3 is available - an imbalance in the speed of read/write operations and brain power, which complicates parallel access to data.

RAID 5. Disk array with parts and divided pairs.

The method is similar to the previous one, but in this new pairing, it is not the entire disk that is visible, but this information is distributed between all the disks. If disk N is selected, then disk N-1 will be available. One service will be seen under any parity, such as RAID 3.4. The smells are not stored on a separate disk, but separated. On the skin disk there is (N-1)/N volume of information and 1/N volume is filled with pairing bits. If there is one disk out of order in the array, then it is lost to waste (the data that was saved on the new one is calculated on the basis of the pairing of other disks “on the lot”). This avoids the need for a computer operator and results in minimal productivity losses (depending on the calculation value of the RAID controller). The advantage is the significantly high speed of reading and writing data, both with great commitments and for a large number of queries. Not enough – the data update is more complex and lower, and RAID 4 has lower readability.

RAID 6. Disk array made of redundant and subdivided pairs.

All the power comes down to the fact that there are two schemes of pairing. The system can hold up to two disks. The main difficulty is those for which it is necessary to perform more operations when recording. Because of this, the liquidity of the entry is extremely low.

Combined (nested) RAID levels.

The fragments of a RAID array are transparent to the OS, and it is time to create arrays, elements such as disks, and arrays of other components. Zazvichay stinks are written with a plus. The first number means those arrays of which level are included as elements, and the second number means those whose organization is the top level, which combines elements.

RAID 0+1

A combination, such as a RAID 1 array, assembled on the basis of a RAID 0 array. As in a RAID 1 array, only half of the disk capacity will be available. However, as in RAID 0, the speed will be greater, even with one disk. To implement such a solution, at least 4 disks are required.

RAID 1+0

Same as RAID 10. With a stripe of mirrors, then a RAID 0 array will be created with RAID 1 arrays. Almost similar to the previous solution.

RAID 0+3

The massif with the visible parity above Cherguvanny. And the array of the 3rd level, in which the blocks are divided into blocks and written to the array RAID 0. Combinations, other than the simplest 0+1 and 1+0, require specialized controllers, often requiring expensive ones. The reliability of this type is lower than that of the current version.

RAID 3+0

Same as RAID 30. With a stripe (RAID 0 array) from RAID 3 arrays. There is a very high data transmission speed, at the same time with poor data transfer capacity. The data is initially divided into blocks (like RAID 0) and distributed into array elements. There they are again divided into blocks, their pairing is taken into account, blocks are written to all disks except one, which is written to the pair. In this case, one can enter one of the disks of the skin array, which is included in the RAID 3 storage.

RAID 5+0 (50)

It is created by combining RAID 5 arrays into a RAID 0 array. There is a high speed of data transfer and query processing. It has medium elasticity and good wash resistance. The combination of RAID 0+5 is also practical, but more theoretical, because it gives little advantage.

RAID 5+1 (51)

Complementary mirroring and drawing with a separated couple. Another option is RAID 15 (1+5). It has very high wash resistance. The 1+5 array works with three disks, and the 5+1 array with five of eight disks.

RAID 6+0 (60)

Cherguvannya with subdivided parity. In other words – stripe with RAID 6. As has already been said, RAID 0+5, RAID 6 with stripes without expansion (0+6). Similar techniques (stripe from arrays with parity) allow you to move the fluidity of the robotic array. Another advantage is that you can easily move things along without complicating the situation due to the delays required to calculate and record a large number of pairs.

RAID 100 (10+0)

RAID 100, also written as RAID 10+0, is a stripe from RAID 10. In its essence, it is similar to the larger RAID 10 array, but contains twice as many disks. This same “tri-topic” structure has its own explanation. Most often, RAID 10 is disabled in hardware, or by the controller, and the stripe is disabled in software. They go to such lengths to eliminate the problems that were initially discussed - controllers are paying attention to scalability and by integrating an increased number of disks into one controller, the increase is not likely to be large for many minds. Software RAID 0 allows you to create it on the basis of two controllers, which are combined on board RAID 10. Thus, we have a unique “dance neck” in a particular controller. Another neat point is to bypass the problem with the maximum number of slots on one controller - double their number and double the number of available slots.

Non-standard RAID modes

Subspecies parity

In addition to the list of RAID levels, there is a parallel pairing, which is implemented and is therefore called a diagonal pairing. The secondary pairing is already implemented in RAID 6. However, when subordinated to it, the pairing is applied to other data blocks. Recently, the RAID 6 specification was expanded, so diagonal parity can be included in RAID 6. For RAID 6, parity is considered as the result of adding modulo 2 bits that go in a row (the sum of the first bit on the first disk, the first bit on the other etc. .), then in the diagonal pair there is displacement. Working in disk failure mode is not recommended (due to the complexity of calculating spent bets from control sums).

With the expansion of the NetApp RAID array with a secondary pairing, it is subject to the updated designation of RAID 6. The Vikorist admin is based on the classic RAID 6 implementation of the data recording scheme. Recording is carried out initially to the NVRAM cache, with a guarantee of uninterrupted life, to avoid wasting data when the electricity is turned on. The controller software, as far as possible, writes more than one whole block to the disk. This scheme provides greater protection than RAID 1 and greater speed of operation than RAID 6.

RAID 1.5

Having been registered by the Highpoint company, it is now stuck even often in RAID 1 controllers, without seeing this particularity. The point comes down to simple optimization - data is written as to the primary RAID 1 array (less than 1.5 in fact), and data is read from two disks (like RAID 0). For a specific implementation of Highpoint, which was installed on DFI boards of the LanParty series on the nForce 2 chipset, the increase will be noticeable, and sometimes zero. This is probably due to the low speed of the controllers of this generator.

Combines RAID 0 and RAID 1. Creates a minimum of three disks. The data is written together on three disks, and a copy of it is written on one disk. If one block is written to three disks, a copy of the first part is written to another disk, and a copy of the other part is written to the third disk. If you have a limited number of disks, it is better to use RAID 10.

When using RAID 5, make sure that one disk is spared, so that if there is a failure, the system immediately begins to rebuild the array. During initial operation, this disk runs idle. A RAID 5E system transfers one disk's memory as an element to the array. This free disk is distributed throughout the entire array and is located at the end of the disks. The minimum number of disks is 4 pieces. The available service is equivalent to n-2, the volume of one disk is selected (being distributed among the system) for parity, the volume of one more disk is large. When the disk goes out of tune, an array of up to 3 disks (with a minimum number) is compressed to fill the free space. A primary RAID 5 array is available, which can last up to one more disk. When a new disk is connected, the array is expanded and occupies all disks again. Warto note that when pressed and unpacked, the disc is not stable until another disc comes out. Also unavailable for reading/writing at this time. The main advantage is the high fluidity of the work, the scraps of cutting are generated on a larger number of disks. The downside is that it is not possible to assign this disk to up to several arrays, which is possible with a simple RAID 5 array.

RAID 5EE

What emerges first is that the areas of free space on the disks are not reserved by one piece at the edge of the disk, but are divided into blocks with pairs of pairs. This technology significantly speeds up renewal after a system failure. Blocks can be written directly to a free space, without the need to move around the disk.

Similar to RAID 5E, there is an additional storage disk to increase the fluidity of the work and the distribution area. More space is shared between other disks and is located at the end of the disks.

This technology is a registered trademark of Storage Computer Corporation. The array is configured on RAID 3, 4, optimized for increased productivity. The main advantage lies in vicoristic caching of read/write operations. Data transfers take place asynchronously. The SCSI drives will need to be installed in the near future. The speed is higher for RAID 3.4 solutions by approximately 1.5-6 times.

Intel Matrix RAID

And the technology presented by Intel at new bridges, starting with ICH6R. The essence comes down to the possibility of combining RAID arrays of different levels on partitions of disks, or even across disks. Let's say, on two disks you can organize two partitions, two of them to save the operating system on a RAID 0 array, the other two - working in RAID 1 mode - to save copies of documents.

Linux MD RAID 10

This is a RAID driver of the Linux kernel, which makes it possible to create a larger version of RAID 10. Thus, for RAID 10 it was necessary to separate a pair of disks, this driver can work with an unpaired one. The principle for three disks will be the same as for RAID 1E, if there is a striping of disks across the same copy and a striping of blocks, as in RAID 0. For four disks, it will be equivalent to the original RAID 10. what can be set in which area A copy is saved on the disk. Let's say the original will be on the first half of the first disk, and the copy will be on the other half of the other. With the other half of the tributes - by chance. Data can be duplicated several times. Saving copies on different parts of the disk allows for greater access speed due to the diversity of the hard drive (access speed changes depending on the data transfer on the platter, so the difference becomes vichi).

Developed by Kaleidescape for use on your media devices. Similar to RAID 4 with vicoristic pairing, but vicoristic is another method of resistance to damage. Users can easily expand the array by simply adding disks, and if they are unable to store data, the data will simply be added to a new location, as needed.

Development of the Sun company. The biggest problem with RAID 5 is the loss of information after connection to life, if the information from the disk cache (which is a non-volatile memory that does not save data without electricity) cannot be saved on a magnetic plate. This dispersion of information in the cache and on disk is called incoherence. The organization of the array itself is related to the Sun Solaris file system – ZFS. The primus record is used instead of the disk cache memory; it is possible to update not only the entire disk, but also a block per block, if the control sum has not been lost. Another important aspect is the ideology of ZFS - it does not change data out of necessity. Then write about updating the data and then, having re-converted to see that the operation has already taken place, change the indicator to them. In this way, it is possible to avoid wasting data during modification. Other files are duplicated instead of creating checksums. The goal is to work with the file system so that you are familiar with the data structure (RAID array) and can see the place for it. This also applies to RAID-Z2, which, like up to RAID 6, is designed to contain two drives with the help of two checksums.

Those that do not have RAID in principle, but often choose to use it at the same time. Literally translated as “just a bunch of disks” The technology combines all the disks installed in the system into one great logical disk. Then, instead of three disks, one great one will be visible. All discs are being reviewed. There is no acceleration, no reliability, no productivity.

Drive Extender

The function is built into Window Home Server. It combines JBOD and RAID 1. If necessary, make a copy, do not duplicate the file, but put the NTFS partition label that indicates the data. With a simple system, the file is copied so that the space on the disks is maximum (disks of different sizes can be used). Allows you to achieve many advantages of RAID - visibility and the ability to easily replace the disk that is in harmony, and its update in the background, visibility into the location of the file (regardless of which disk vin knows). It is also possible to conduct parallel access from different disks using additional labels, resulting in performance similar to RAID 0.

Split by Lime technology LLC. This scheme differs from basic RAID arrays in that it allows you to mix SATA and PATA drives in one array and drives of different types and speeds. For the check sum (parity) the vision disk is used. Data does not compete between discs. When you use one disk, only the files that are saved on that disk are lost. However, due to additional steaminess, the stench may be renewed. UNRAID was introduced as an extension to Linux MD (multidisk).

Most types of RAID arrays have not been expanded; some are being abandoned in narrow areas of stagnation. The most widespread, from simple servers to cob servers, are RAID 0, 1, 0+1/10, 5 and 6. Whatever raid array you need for your tasks is up to you. Now you know why their roles are the same.

The saying “As long as you don’t hit the grim, a man can’t cross each other” is known to most people. Zhitteva won: as long as this other problem doesn’t run into the koristuvach, he won’t even think about it. The life block has died and a couple of devices have come with it - the cook rushes to read articles on a similar topic about tasty and healthy food. When the processor burns out or starts to fail due to overheating, “Selected” will show a couple of messages on various forums where CPU cooling is discussed.

With hard drives it’s the same story: the sweet worm’s screw, having cracked its heads goodbye, deprives our fleeting light, the ruler of the PC begins to rush around to ensure the expansion of the accumulator’s living minds. However, the coolest cooler cannot guarantee a long and happy life for the disk. There are a lot of officials involved in the service of the accumulator: both a sluice in the production plant, and a tapping of the body with a foot (especially since the body is standing here on the underside), and a drink that has passed through the filter, and a high-voltage power supply. , sent by the life block... There is only one output - reserve copying information, and if you need backup on the go, then you will need a RAID array, since today the motherboard may have some kind of RAID controller.

At this point, we begin with a short excursion into the history and theory of RAID arrays. The abbreviation RAID itself stands for Redundant Array of Independent Disks. Previously, instead of independent, inexpensive (inexpensive) lived, but later this designation lost its relevance: all disk drives became inexpensive.

The history of RAID began in 1987, when the article “Enclosure for over-the-air arrays with low-cost disks (RAID)” was published, signed by comrades Peterson, Gibson and Katz. The note described the technology of combining multiple primary array disks to extract more fluid and reliable storage. The authors of the material also informed readers about the number of types of arrays – from RAID-1 to RAID-5. A year before the descriptions, about twenty years ago, the arrays received RAID arrays of zero level, and gained popularity. So what is all this about RAID-x? What is its essence? Why are the stench called supermundane? Whom we are trying to get back to you.

To put it simply, RAID is a thing that allows the operating system not to know how many disks are installed in the computer. Combining hard drives in RAID arrays is a process directly similar to dividing a single space onto logical drives: we form one logical drive based on several physical ones. In order to do this, we will need either reliable software (we won’t talk about this option - it’s unnecessary), or a RAID controller installed in the motherboard, or a PCI or PCI Express device that is inserted into the slot. The controller itself puts the disks into an array, and the operating system no longer works with the HDD, but with the controller, which does not report anything unnecessary to it. And the axis of options is combining many disks into one complete disk, more precisely, close to ten.

What is RAID?

The simplest of them is JBOD (Just a Bunch of Disks). Two hard drives are glued into one sequentially, information is written first to one, and then to another disk without breaking into scraps or blocks. For two 200 GB drives, we only need one for 400 GB, which works practically as well, but in reality, with a fraction of the speed of two disks.

JBOD - let's add a zero level array, RAID-0. There is also another variant of the name of the array of this level - stripe (black), otherwise called - Striped Disk Array without Fault Tolerance. This option also transfers the combination of n disks into one volume, increasing n times, but the disks are not connected sequentially, but in parallel, and information is written to them in blocks (the volume of the block is determined by the user's form baths to the RAID array ).

If you have two drives that go up to a RAID-0 array, you need to write down the sequence of digits 123456, the controller will divide this drive into two parts - 123 and 456 - and write it first on one disk, and on the other - on the other. Each disk can transfer data... well, at a speed of 50 MB/s, and the total speed of two disks, data from which are taken in parallel, becomes 100 MB/s. Thus, the speed of the robot with data must increase n times (in fact, the increase in speed is less, since the amount spent on searching for data and transmitting it over the bus is not consumed). However, this growth does not happen like this: if even one disk breaks, the information from the whole array is wasted.

RAID array of zero level. The data is divided into blocks and scattered across disks. There is no parity control and redundancy.

There is no superciliousness and no reserve at all. It is possible to use this array as a RAID array only cleverly, but it is even more popular. Few people think about reliability, and even if you don’t rely on benchmarks, you still count megabytes per second. It's not bad, it's not good, it's just a reality. Below we will talk about how to catch fish and preserve hope. RAID-0 upgrade after failure

Before speaking, the additional minus of the stripe array is due to his intolerance. I do not dare to admit that it is wrong to endure any species of hedgehogs or, for example, rulers. No matter what, or even moving the massif itself is a problem. However, if you pull the drives and drivers of the controller to each other, it is not a fact that they appear as one mass and can be quickly processed with the data. Moreover, there are problems when simply connecting (without writing down what!) stripe disks to a “non-ready” (which depends on the type on which the array was formed) controller led to the loss of information in the array. We don’t know how urgent this problem is now, with the appearance of current controllers, but still for the sake of being careful.

RAID array of the first level with four disks. The disks are divided into bets; however, new data is stored on accumulators in the middle of the bet.

The first, as a matter of fact, “overkill” array (and the first RAID that came into being) is RAID-1. Another name - mirror (mirror) - explains the principle of operation: all inputs to the disk array are divided into pairs, and the information is read and written directly to the disks. It turns out that the skin from the disks of the massif contains an exact copy. In such a system, not only the reliability of saving data increases, but also the speed of reading it (you can read it from two hard drives at once), although the speed of writing is lost as much as in one storage device.

As you might guess, I owed such a massive amount of debt to all the Winchesters that I had to enter before. The downside of this solution is that you need twice as much hard drive. However, the reliability of this array is actually not comparable to the reliability of a single disk, but is richer for the price. The exit from the harmony of two Winchesters stretches... well, let's say, it's very small, as it doesn't interfere with the right, for example, the life block. At the same time, be it a sensible person, having found that one disk is in good order with a couple of people, immediately replace it, and immediately after which you end up with another disk, the information will not go anywhere.

As you can see, both RAID-0 and RAID-1 have their shortcomings. How would you wake them up? If you have the smallest amount of hard drives, you can create a RAID 0+1 configuration. For this array, RAID-1 is merged into the RAID-0 array. Or, on the other hand, they create a RAID-1 array from several RAID-0 arrays (the output is RAID-10, the only advantage of which is a shorter data update hour when one disk is out of order).

The reliability of such a configuration with four hard drives is the same as the reliability of a RAID-1 array, and the reliability is actually the same as that of a RAID-0 (in fact, it is higher than everything, but it will be a little lower due to the limited capacity of the controller). In this case, the one-time failure of two disks does not always mean a repeated loss of information: it becomes impossible for the disks to break so that they can only contain the same data, which is unlikely. If the disks are divided into pairs 1-2 and 3-4 and paired into a RAID-0 array, then an immediate failure of disks 1 and 2 or 3 and 4 will lead to the loss of data, as in the case of the immediate death of the first one. the third, the second and the fourth, the first and the fourth, or the second and the third Winchester tribute to be lost intact and safe.

However, the main drawback of RAID-10 is the high disc quality. You can’t tell the price of several (at least!) small Winchesters, especially since only two of them are actually available to us (few people think about reliability and about how much they cost, as we have already said). Signs are given of the great (100%) supernatural nature of saving data. All this has led to the continued popularity of an array option called RAID-5. For its implementation, three disks are required. In addition to the information itself, the controller stores massive parity control blocks on storage devices.

We will not go into detail about the parity control algorithm, except that it allows you to update the parity data and live data from other disks every time you lose information on one of the disks. Blok Parnosti MAM ONE FILICAL DISK I RIVENNALLY RASELLEN ON INIKH INCHECHERS OF THE SYSTEM, INTED IN THE INITY DISK INDITIONAL INFORMATION OF THE DOBOOGE OF THE DO PAILSTISTIS ON ININSH DASKA. Information is divided into large blocks and written to disks across the system, using the 12-34-56 principle in conjunction with a tri-disk array.

Apparently, the best solution for such an array is all the disks minus the capacity of one of them. Updated data, obviously, is not a mitigation, but such a system has high productivity and a reserve of reliability with minimal power (for an array with a volume of 1000 GB, six disks of 200 are required GB). However, the productivity of such an array will still be lower than the flexibility of the stripe system: during each recording operation, the controller will need to update another parity index.

RAID-0, RAID-1 and RAID 0+1, as well as RAID-5 – these levels most often draw on the capabilities of desktop RAID controllers. Much more is available only to folding systems based on SCSI hard drives. However, lucky owners of SATA controllers with the support of Matrix RAID (such controllers are available in the original ICH6R and ICH7R bridges from Intel) can quickly use arrays of RAID-0 and RAID-1, with only two disks, and even then I pay for ICH7R, can combine RAID-5 and RAID-0, as they may be a problem for new storage devices.

How can this be implemented practically? Let's take a look at the simplest case: RAID-0 and RAID-1. Let's say you bought two 400 GB hard drives. You split the storage media into logical disks with a capacity of 100 GB and 300 GB. After using the Intel Application Accelerator RAID Option ROM utility built into the BIOS, you will connect the 100 GB partitions to the stripe array (RAID-0), and the 300 GB partitions to the Mirror array (RAID-1). Now, on a 200 GB flexible disk, you can store, say, games, video materials and other data that require high fluidity of the disk subsystem and are not so important before (that is, about the loss of such and you won’t misbehave any more), but on Mirroring a 300-gigabyte disk moves work documents, email archives, service software and other vital files. When you go out of order on one disk, you add what was placed on the stripe array, and the data you placed on another logical disk are duplicated on the storage that is left over.

The combination of RAID-5 and RAID-0 is based on the fact that part of each disk is placed under a flexible stripe array, and the other part (at least 300 GB per disk) falls into data blocks and parity blocks. then You will select one super-wide disk with a volume of 400 GB (4 x 100 GB) and one smaller disk with a volume of 900 GB (4 x 300 GB minus 300 GB for parity blocks).

As you can see, this technology is extremely promising, and it is not bad that other chipset and controller manufacturers support it. It’s already very exciting to have mothers on two disks of the massif of different regions, swedish and nadiyna.

Axis, perhaps, and all types of RAID arrays that are used in home systems. However, in real life you may have RAID-2, 3, 4, 6 and 7. So let’s see what kind of equals they are.

RAID-2. In an array of this type, the disks are divided into two groups - for data and correction codes, and since the data is stored on n disks, n-1 disks are required to accumulate correction codes. Data is recorded on separate hard drives just like RAID-0; they are divided into small blocks on a number of disks used for storing information. Disks that have been lost save error correction codes, so that you can update the information every time you leave any hard drive. The Hamming method has long been stagnant in ECC-type memory and allows one to correct fractional one-bit data, which can lead to a breakdown, and if two bits are transmitted per data, this will be detected again by additional pair control systems sti. However, no one wanted to trim this bulky structure along with the suspended number of disks, and this type of mass without reducing the width.

Massive structure RAID-3 like this: in an array with n disks, the data is divided into blocks of 1 byte in size and distributed across n-1 disks, and another disk is rotated to save pairs. RAID-2 had n-1 disks for this purpose, and most of the information on these disks was recovered only to correct errors on a large scale, and for simple updating in case of disk failure, add a smaller amount And one visible Winchester stands up.

Level 3 RAID with edged disk for storing parity information. There is no reserve, but the data can be updated.

Apparently, the advantages of RAID-3 compared to RAID-2 are obvious: the impossibility of correcting damages on a large scale and less overpowering. The advantages are as follows: the speed of reading and writing data is high, and to create an array you need very few disks, no more than three. However, this mass is only good for single-tasking work with large files, but you should be wary of problems with speed when frequently refilling small data.

The fifth level array is subdivided into RAID-3 so that the parity blocks are sequentially distributed across all disks in the array.

RAID-4 similar to RAID-3, but differs in that the data is divided into blocks rather than bytes. In this way, it was possible to “overcome” the problem of low speed of data transfer for small businesses. Recording is carried out entirely through those pairings for the block that are generated during recording and recorded on a single disk. It is very rare for this type of vikorist to appear.

RAID-6- not the same RAID-5, but now two blocks of parity are saved on the disk array. Thus, when the two disks come out in harmony, the information can still be updated. Apparently, the increase in reliability led to a change in the core volume of disks and an increase in their minimum capacity: now, apparently, n disks in the array have a minimum volume available for recording data, up to nyuvatime obsyag one disk, multiplied by n-2. The need to calculate two checksums means another disadvantage of RAID-6 versus RAID-5 - low data recording speed.

RAID-7- The trademark of the company Storage Computer Corporation is registered. The structure of the array is as follows: data is stored on n-1 disks, one disk is used to create parity blocks. But there were a number of important details that needed to be sorted out of the bulk of arrays of this type: the data cache and the flexible controller that manages the processing of queries. This made it possible to change the number of disks used to calculate the control sum of the data. As a result, it was possible to significantly increase the speed of data processing (perhaps five or more times).

An array of RAID 0+1 level, or a design of two RAID-1 arrays united in RAID-0. Reliable, smart, expensive.

There are new shortcomings: the high level of implementation of such an array, the complexity of its maintenance, the need for an uninterruptible service to prevent the loss of data from the cache memory during storage interruptions. You’re unlikely to get a message from this type of mass, but if you get a rap here, write to us, and we’ll be amazed at the new one.

Creation of the massif

I’m sure you’re already screwed with the choice of massif type. Since the board has a RAID controller, you don’t need anything other than the required number of disks and drivers for the controller itself. Before speaking, please remember: it is sense to combine only the same discs into an array, and moreover, for one model. With disks of different volumes, you can use a controller, and, above all, you can use only part of a large disk, which is similar to smaller disks. In addition, the speed of the stripe array depends on the speed of the maximum disk drive. And my pleasure to you: do not try to create a fancy RAID array. It is possible that if you have any problems with the system, it will not be easy for you, and the renewal of productivity will be even more difficult. In addition, it is not safe to place a number of systems on such an array: even all the programs that are responsible for choosing an OS, enter information from the service areas of the hard drive and, obviously, destroy the array. It’s better to choose a different scheme: one disk is the original one, the others are combined into an array.

Matrix RAID at dii. Part of the disk volume is taken up by the RAID-0 array, and part of the space that is lost is taken by the RAID-1 array.

The RAID array starts from the BIOS of the RAID controller. Others (only those with integrated controllers, but not always) are included in the main BIOS of the motherboard, and sometimes they are activated after self-testing, otherwise you may need them there. The BIOS itself sets the required parameters of the array, as well as the size of data blocks, selected hard drives, and so on. After all, it is enough to save the settings, exit the BIOS and return to the operating system.

There you will definitely need to install the controller drivers (as a rule, a floppy disk with them is delivered to the motherboard or to the controller itself, otherwise they can be recorded on a disk with other drivers and service software), re-install, and that’s it, everything is ready before work. You can split it into logical disks, format it and store it with data. Just remember that RAID is not a panacea. It protects you from wasting data when the hard drive dies and minimizes the consequences of this result, rather than avoiding voltage surges and failures of the unknown life block, which causes damage to the drives, without regard to its “massiveness”.

Failure to carefully set the temperature regime of the disks can significantly shorten the life of the HDD, as it happens, all disks in the array come out in good order, and all data is irrevocably wasted. Today, modern hard drives (especially IBM and Hitachi) are very sensitive to the +12 V channel and do not like to send the smallest voltage change to the new one, so before purchasing all the equipment necessary for energizing the array, check the The same voltage and if necessary, turn on the new power supply before shopping list.

The life of hard drives, as well as all other components, in another life block, at first glance, is simple, but there are a lot of pitfalls in such a life scheme, and you need to think a hundred times before relying on such croc. With refrigeration, everything is simpler: you just need to ensure that all the hard drives are ventilated, plus don’t put them in the same room. Simple rules, but, unfortunately, not all of them are followed. And attacks, when the disks in the array die overnight, are not uncommon.

In addition, RAID eliminates the need to regularly back up data. Mirror mirror, if you simply zip or erase files, another disk will not help you in any way. So, make sure to back up as soon as you can. This rule applies regardless of the presence of RAID arrays in the middle of the PC.

So, are you RAIDy? So? Miracle! Just in the pursuit of volume and speed, don’t forget the following saying: “I’ll start praying to God, and my forehead will break.” Good luck with your disks and reliable controllers!

The Cost Benefit of Noisy RAID

RAID is good news without regard for pennies. But let's take a look at the price of the simplest stripe array with a capacity of 400 GB. Two Seagate Barracuda SATA 7200.8 drives of 200 GB each will cost you approximately $230. RAID controllers are built into most motherboards, so we can remove them without costs.

For the same model, a 400 GB drive costs $280. The price is $50, and with that little money you can get an extra block of life, which you will definitely need. I’m no longer talking about those that the productivity of a folded “disk” at a lower price will be twice as much as the productivity of one hard drive.

Let’s now carry out the cleaning, focusing on the storage space of 250 GB. There aren’t enough 125 GB disks, so we’ll take two 120 GB hard drives. The price of a skin disk is $90, the price of one 250 GB hard drive is $130. Well, for such obligations you have to pay for productivity. How about taking a 300 GB array? Two 160 GB disks - approximately $200, one 300 GB - $170... I know not the same. It turns out that RAID is only effective when the drives are removed, which is a great benefit.