A group/redundant array of independent disks1 or RAID (the English redundant array of independent disks) refers to a data storage system that uses multiple data storage units (hard disks or SSDS) among those distributed or replicated data.
Depending on your configuration (often referred to as level), the benefits of a RAID against a single hard drive are one or more of the following: greater integrity, tolerance to failures, transfer rate and capacity. In their implementations original, its key advantage was the ability to combine multiple low-cost devices and older technology in a set that offered greater capacity, reliability, speed, or a combination of these that a single device of the latest generation and highest cost.
At the very simplest level, RAID combines multiple hard drives into a single logical unit. So, instead of seeing several different hard drives, the operating system sees only one. The RAID is typically used on servers, and is usually (although not necessary) are implemented with disk drives of the same capacity. Due to the decrease in the price of the hard drives and the increased availability of RAID options included in the chipsets of the motherboards, the RAID is also found as an option on the personal computers more advanced. This is especially common in computers dedicated to intensive tasks, and that requires ensuring the integrity of the data in case of system failure. This feature is available in the RAID systems by hardware(depending on which structure we choose). By contrast, systems based on software are much more flexible and the hardware-based add a point of failure more to the system (the RAID controller).
All implementations may support the use of one or more spare disks (hot spare), drives pre-installed that can be used immediately (and almost always automatically) after the failure of one disk in the RAID. This reduces the time period of repair to shorten the reconstruction time of RAID.
RAID levels standard
The RAID levels most commonly used are:
• RAID 0: striped Array.
• RAID 1-Set in mirror.
• RAID 5: striped Array with distributed parity.
RAID 0 (Data Striping, Striped Volume)
A RAID 0 (also called a striped array, volume, split, striped volume distributes the data evenly between two or more disks (usually occupies the same space on two or more disks) without parity information to provide redundancy. It is important to note that the RAID 0 was not one of the RAID levels the original and that is not redundant. The RAID 0 is normally used to provide a high write performance because data is written to two or more disks in parallel, even if the same file is only present once in the set. RAID 0 can also be used as a way of creating a small number of large virtual disks from a large number of small physical disks. A RAID 0 can be created with disks of different sizes, but the storage space added to the set will be limited by the size of the smallest disk (for example, if a disk 450GB is divided with one of 100 GB, the size of the resulting set will be only 200 GB, because each disk provides 100 GB). A good implementation of a RAID 0 will split the read and write operations in blocks of equal size, so that it will distribute the information evenly between the two disks. It is also possible to create a RAID 0 with more than two disks, although, the reliability of the whole will be equal to the average reliability of each disk between the number of discs in the set; that is to say, the reliability —measured as MTTF or MTBF— it is (approximately) inversely proportional to the number of disks in the array (as to which the joint fails, it is sufficient to do any of your disks). Not to be confused with RAID 0 with a Spanned Volume (Spanned Volume) in which you add multiple spaces will not be used for several drives to form a single virtual disk. It may be that in a Spanned Volume, the file to retrieve is present in a single disc of the set, because here there is equitable distribution of the data (as we said for RAID 0), so in that case it would not be possible to recover data parallel and we would not have to improve the performance of reading.
RAID 1 (mirror)
A RAID 1 creates an exact copy (or mirror) of a set of data on two or more disks. This is useful when we want to have more security in light of capacity, because if we lose a disk, we have the other one with the same information. A RAID 1 set can only be as large as the smallest of your disk drives. A RAID 1 classic consists of two disks in mirror, which exponentially increases the reliability over a single disk;that is to say, the probability of failure of the whole is equal to the product of the probabilities of failure of each one of the discs (thus for which the joint fails it is necessary to do soall of their albums).
Additionally, since all the data is in two or more disks, with hardware usually independent, the read performance increases approximately as a multiple linear in the number of copies; that is to say, a RAID 1 can be simultaneously reading two different data on two different disks, so its performance is doubled. To maximize the benefits on the performance of the RAID 1 is recommended the use of disk controllers are independent, one for each disk (which some referred to as splitting or duplexing).
As in the RAID 0, the average reading time is reduced, since the sectors to look for can be divided between the disks, decreasing the search time and upping the transfer rate, with the only limit of the speed supported by the RAID controller. However, many RAID cards 1 IDE old read-only one disk in the pair, so its performance is equal to that of a single disk. Some implementations of RAID 1 old also read from both disks simultaneously and compare the data to detect errors.
When writing, the set behaves like a single disk, because the data must be written to the disks in the RAID 1. Therefore, the write performance does not improve.
The RAID 1 has many advantages in administration. For example, in some environments 24/7, it is possible to «split the mirror»: to mark a disk as inactive, do a backupof that disk, then «rebuild» the mirror. This requires that the application set management support data recovery of the disk at the time of the division. This procedure is less critical than the presence of a feature snapshot in some file systems, in which reserves some space for changes, presenting a static view at a point in time because of the file system. Alternatively, a set of discs can be stored in a similar way as with the traditional tapes.
A RAID 5 (also called distributed parity) is a division of data at the level of blocks , which distributes information and parity across all disks members of the set. The RAID 5 has achieved popularity thanks to its low cost of redundancy. Usually, RAID 5 is implemented with hardware support for the calculation of the parity. RAID 5 will need a minimum of 3 disks to be implemented.
In the graphic example above, a read request for block «A1» would be serviced by disk 0. A request for simultaneous reading of the block «B1» would have to wait, but a read request for B2 could be addressed concurrently as they would be served by the disk 1.
Each time a block of data is written to a RAID 5, generates a parity block within the same slice (stripe). A block is comprised often of many consecutive sectors of a disk. A series of blocks (one block of each of the disks in the array) receive the collective name of division (stripe). If another block, or some portion of a block, is written in that same division, the parity block (or a part thereof) is recalculated and back to writing. The disk used for the parity block is staggered from one division to the next, hence the term «distributed parity blocks». The scriptures in a RAID 5 are costly in terms of disk operations and traffic between the disks and the controller.
The parity blocks are not read in read operations of data, as this would be an unnecessary overhead and would diminish performance. However, the parity blocks are read when the reading of a data sector results in an error of CRC. In this case, the sector in the same relative position within each of the blocks of data remaining in the division and within the parity block in the division are used to reconstruct the sector wrong. The CRC error is hidden as the rest of the system. In the same way, if a disk fails the array, the parity blocks of the remaining disks are combined mathematically with the data blocks of the remaining disks to reconstruct data from the failed disk «on the fly».
The above is sometimes referred to as Mode Interim Recovery of Data (Interim Data Recovery Mode). The system knows that a disk has failed, but only for the purpose of the operating system you can notify the administrator that a drive needs to be replaced: running applications continue to operate beyond the fault. The reads and writes to continue normally in the set of disks, albeit with some degradation of performance.The difference between RAID 4 and RAID 5 is that, in the Internal Mode Data Recovery, RAID 5 might be slightly faster, because, when the CRC and parity are in the disk that failed, the calculation does not have to be performed, while in RAID 4, if one of the data disks fails, the calculations have to be performed at each access.
The failure of a second disk causes the complete loss of data.
The maximum number of disks in a redundancy group of the RAID 5 is theoretically unlimited, but in practice it is common to limit the number of units. The disadvantages of using groups of redundancy are older, they are in a higher probability of simultaneous failure of two disks, a higher reconstruction time and a higher probability of finding a sector unrecoverable during a rebuild. To the extent that the number of disks in a RAID set 5 grows, the MTBF (mean time between failures) can be lower than that of a single disk. This happens when the probability of failure of a second disk in the N-1 remaining disks of a set that has failed a disk in the time it takes to detect, replace and recreate that disk is greater than the probability of failure of a single disk. An alternative that provides protection parity dual, thus allowing a greater number of disks per group is RAID 6.
Some vendors RAID avoid mount disks of the same batch in a redundancy group to minimize the probability of simultaneous failures at the beginning and the end of its useful life.
Implementations of RAID 5 have poor performance when subjected to workload which includes many writes smaller than the size of a division (stripe). This is because parity must be updated for each write, requiring you to perform sequences of read, modification and write both the data block and the parity. More complex implementations often include caches write non-volatile to reduce this performance problem.
In the case of a failure of the system when there are scriptures that are active, the parity of a division (stripe) may be left in an inconsistent state with the data. If this is not detected and repaired before a disk or block fails, data loss may occur due to the use of a parity incorrect to reconstruct the lost block in the division. This potential vulnerability is sometimes known as a «hole writing». Are common the use of cache is not volatile and other techniques to reduce the probability of occurrence of this vulnerability.
HOW TO: create a disk (RAID 0)
A sectioned volume (RAID 0) combines free space areas of several hard disks (between 2 and 32) into a single logical volume. Data written on a sectioned volume is interleaved on all disks at the same time instead of sequentially. As a consequence, the disk performance will be faster in a RAID 0 volume compared to another type of disk configuration. Administrators prefer to use sectioned volumes when the speed of entry and exit is important. Any file system in a sectioned volume can be used, including FAT, FAT32 or NTFS.
- There must be at least two hard drives (IDE, SCSI, or mixed architectures are supported).
- All disks involved in the sectioned volume must be dynamic disks.
- Each part of the free space must be identical (for example, in terms of size and in terms of the type of file system).
How to configure the disk management system
- Click the right- mouse (mouse) in My Computer and then click Manage .
- Click the plus sign (+) next to Storage to open the Storage console tree.
- Click on the Disk Management folder .
- On the View menu , select Up and click Disk List . In the right panel, a column with the attributes of each of the system’s disks is shown.
- On the View menu , select Down and click Graphical View .
A graphic view with color code of the system disks will be displayed.
The disk description panel (shown in gray) is in the left pane of the volume description, which is displayed in color. The disk description contains information about the disk number, the basic or dynamic configuration, the size and the online or offline status of each of the disks.
The descriptions of volumes are color coded. They contain information about each of the volumes, such as the letter of the unit (if it has one assigned), whether they are assigned or not, the partition or size of the volume and its status.
Requirements to ensure that the disks are configured to support a sectioned volume
- Discs : At least two discs are required to support the creation of bands.
- Write : all the discs involved in the creation of bands must be dynamic. The conversion from basic to dynamic is done quickly without loss of data. After the conversion process is complete, restart the computer.
- Capacity : The sectioned volume can occupy the whole disk or only 50 megabytes (MB) of each disk.
- Unassigned space : All the disks that you want to update to dynamic disk must contain at least 1 MB of free space at the end of the disk in order for the update to succeed. Disk Management automatically reserves this free space when it creates partitions or volumes on a disk, but this space may not be available on disks with partitions or volumes created by other operating systems.
- State : All disks involved in a sectioned volume must be online when creating the sectioned volume.
- Device type : You can create bands on any dynamic disk even if there is a mixture of unit architectures in the system. For example, the IDE, EIDE and SCSI units can be used in a sectioned volume.
How to upgrade to dynamic disks
If the discs that will participate in the sectioned volume are already dynamic, go to the next section («How to convert to a sectioned volume»).
NOTE : You must be logged on as an administrator or as a member of the Administrators group to complete this procedure. If the computer is connected to a network, the network policy setting may prevent you from completing this procedure.
To update a basic disk to a dynamic disk:
- Before updating discs, close all programs that are running on these discs.
- Click the right- mouse (mouse) into the gray disk description panel to the left of the panels color-coded volume, and then click Upgrade to Dynamic Disk .
- If the second disk is not a dynamic disk, follow the previous steps to update it to a dynamic disk.
How to perform the conversion to a sectioned volume
In this scenario, there are two disks in the computer, Disk 0 and Disk 1. Both are dynamic and each has at least 1 gigabyte (GB) of unallocated free space for a total volume of 2 GB.
- In the lower right pane of the Disk Management tool, right – click button on the mouse (mouse) in unallocated free space on either disk, and then click Create Volume .
- When the Create Volume Wizard appears, click Next .
- Click Sectioned Volume , Volume Type, and Next .
- In the left panel, in
Select two or more discs,
a list of all discs that have sufficient unallocated free space to participate in the sectioned volume will appear.
In the right panel, in selected Dynamic Disks , the disk you clicked with the right mouse button in step one will be displayed .
- In the left pane, under All available dynamic disks , click the disk, and then click Add .
All the disks shown in the right panel will have the label «Dynamic disks selected». Look in the lower part of the Select Diskdialog box , on the Size tab . The box for all selected disks shows the maximum sectioned volume size that you can adjust.
NOTE: The volume of each disc is the same size in the entire sectioned volume. For example, if you have 100 MB on the first disk, you have 100 MB on the second, etc. Therefore, the total size of our combined volumes is twice the minimum size of the volumes in the two disks.
You can reduce the volume size from the maximum so that the wizard displays automatically. To do this, click on the arrows in the Disk Size box to reduce the volume size of this disk. Note that, in a system with two disks, the total size of sectioned volume is double the inserted size. The Total Volume Size box , in the right pane, shows the actual size of the sectioned volume.
- Click Next to advance to the Assign drive letter or wizard path page .
- You may now want to assign a drive letter to the sectioned volume (you can also do it later). To do this, click Assign drive letter,and then insert an available drive letter.
Alternatively, you can click
Do not assign a drive letter or path
. You can also click
Mount this volume on an empty folder that supports drive paths:
. However, this selection goes beyond the objective of this article.
- Once you have inserted a drive letter for the sectioned volume, click Next .
Format this partition with the following settings
, and then follow these steps:
- Insert an acceptable file system type such as FAT32 or NTFS.
- Leave the default selection in the Size box of the allocation unit .
- In the Volume Label box , you can keep the «New vol» label, or you can write your own label.
- Now you can click to select the Quick Format checkbox and
Enable file and folder compression
. You can delay both tasks if you wish.
- Click Next , make your selection in the Summary window, and then click Finish .
The sectioned volumes are displayed on the two disks of the system. They have the same color code, the same unit letter (if you assigned the unit during the procedure) and have the same size.
- Do not mix hardware RAID-0 with software RAID-0.
- A sectioned volume can not maintain the system or start the partition of a Windows 2000-based system.
- It can not extend or reflect sectioned volumes.
- The sectioned volumes lack fault tolerance. This means that if one of the discs is damaged or stops working correctly, it will be completely lost.