Post by Javier on Apr 9, 2013 13:09:23 GMT
The subject is complex enough to fill many pages but here is a short summary.
RAID (Redundant Array of Inexpensive Disks) is a clever way to achieve higher transfer speeds, capacity and hardware failure protection using two or more HDDs. RAID does not protect against user errors, for example if you delete/overwrite a file by mistake.
There are several types of RAID: 0, 1, 0+1 (aka 10), 5, 6, 50, 60, etc. Each type has its advantages and disadvantages.
RAID can be performed by the OS (free) or by a dedicated hardware disk controller (can range from affordable to quite expensive for enterprise class models). OS based RAID usually is limited to 0, 1 and 5 though there apps that allow for other types.
What is RAID 0 (aka stripping)?
Minimum HDDs required = 2.
Hardware failure protection = none
Speed = fastest
Capacity = maximum
This type of RAID increases transfer speeds at the cost of a little increase in access time. Using two identical HDDs transfer speed will tend to double, the more HDDs the higher the speed until the bus maximum real speed is reached, from that point further disks won't improve and can actually degrade performance. On the downside there is a complete lack of hardware protection, there is no data redundancy so if one drive in the array fails the whole volume is lost and probably for good.
What is RAID 1 (aka mirroring)?
Minimum HDDs required = 2.
Hardware failure protection = very high
Speed = slow
Capacity = 50% wasted space
The main advantage of this type of RAID is strong hardware protection against data loss due to a HDD failure. One disks acts as master the other as salve and all operations carried on master are cloned to slave. If master dies, slaves has an identical copy of its content so 0 loss. The surviving disk can be used as a standalone disk on the same or a different computer without any problem.
What is RAID 0+1 aka RAID 10?
Minimum HDDs required = 4.
Hardware failure protection = high
Speed = very fast
Capacity = 50% wasted space
This type is a combination of the first two types. the first pair of disks act in RAID 0 mode providing high throughput. The second pair, also in RAID 0 mode, act as a mirror of the first providing redundancy. The number of disks in the array can be increased always in pairs to gain both speed and capacity up to the maximum drives supported by the mainboard or used disk controller.
What is RAID 5?
Minimum HDDs required = 3.
Hardware failure protection = high
Speed = very fast reading, mid to fast writting
Capacity = 1 disk in the array is lost to data parity storage. The more drives in the array the less wasted space.
In this type one disk stores redundant data and the other two store 50% of the data (3 disk array). When reading from disk it performs as a RAID 0 array providing the best performance but when writing, the OS or the controller have to perform data parity calculations reducing throughput performance. How much performance is lost? It depends on available CPU power of the PC (OS RAID) or of the hardware RAID controller's processor. Provides good protection against hardware failure (as long as only one drive fails...) but depending on drive size and RAID system (software Vs hardware, high end Vs low end) array rebuilding times can get quite lengthy. In the unusual case of two consecutive drives failing before array rebuilding is finished chances of data recovery can get quite slim.
What is RAID 6?
Minimum HDDs required = 4.
Hardware failure protection = very high
Speed = very fast reading, mid to fast writing
Capacity = 2 disk in the array are lost to data parity storage. The more drives in the array the less wasted space.
Identical to RAID 5 except that instead of one drive used for parity two are used doubling the hardware failure tolerance. This type is usually only provided by add on hardware RAID disk controllers. Is not common in home environments due to higher costs.
What are RAID 50 or RAID 60?
They are two types of RAID arrays are usually only found on big size enterprise storage systems. By combing RAID 5 arrays of RAID 0 arrays They provide extreme speeds and fault tolerance but they are not within average Joe's budget. If SAS controllers with port expanders are used a RAID 50 or 60 can comprise hundreds of HDDs. Special rack mounted cabinets are used with hot swappable trays and redundant power supplies and cooling systems.
Important options to consider when working with RAID arrays.
Hot spare disks.
Some controllers or OS RAID apps (including Intel's on-board RAID) offer the possibility of designating a disk as a "Hot Spare". These disks cannot be used for anything else, have to stay unpartitioned or formated and have to be of at least the same size as the biggest disk in the array. When a disk in an array fails the OS or the RAID controller automatically uses this "Hot Spare" to start rebuilding the broken array in the background. This is not only convenient but as it acts as soon as the failure is detected it reduces the chances of data loss due to another disk failing.
High end controllers allow for both "dedicated" and "general" hot spares disks, this means you can assign a hot spare to an specific array in case you have arrays for different purposes (IE an array of fast disks of small size and an array of large capacity but slower disks).
Hot swap disk bays.
They are cases for mounting HDDs in a way that makes it very quick and easy to remove them from the computer, replace the disk inside and put them back in the computer. The are probably only relevant on systems running 24/7 with some criticity (IE. enterprise servers).
Copy Back protection.
After a disk failure and the RAID controller has used the available "Hot Spare" disk to rebuild the array, when the user replaces the failed disk with a new one, Copy Back restores the system to the fresh drive and the "Hot Spare" disk recovers its protective status again.
RAID (Redundant Array of Inexpensive Disks) is a clever way to achieve higher transfer speeds, capacity and hardware failure protection using two or more HDDs. RAID does not protect against user errors, for example if you delete/overwrite a file by mistake.
There are several types of RAID: 0, 1, 0+1 (aka 10), 5, 6, 50, 60, etc. Each type has its advantages and disadvantages.
RAID can be performed by the OS (free) or by a dedicated hardware disk controller (can range from affordable to quite expensive for enterprise class models). OS based RAID usually is limited to 0, 1 and 5 though there apps that allow for other types.
What is RAID 0 (aka stripping)?
Minimum HDDs required = 2.
Hardware failure protection = none
Speed = fastest
Capacity = maximum
This type of RAID increases transfer speeds at the cost of a little increase in access time. Using two identical HDDs transfer speed will tend to double, the more HDDs the higher the speed until the bus maximum real speed is reached, from that point further disks won't improve and can actually degrade performance. On the downside there is a complete lack of hardware protection, there is no data redundancy so if one drive in the array fails the whole volume is lost and probably for good.
What is RAID 1 (aka mirroring)?
Minimum HDDs required = 2.
Hardware failure protection = very high
Speed = slow
Capacity = 50% wasted space
The main advantage of this type of RAID is strong hardware protection against data loss due to a HDD failure. One disks acts as master the other as salve and all operations carried on master are cloned to slave. If master dies, slaves has an identical copy of its content so 0 loss. The surviving disk can be used as a standalone disk on the same or a different computer without any problem.
What is RAID 0+1 aka RAID 10?
Minimum HDDs required = 4.
Hardware failure protection = high
Speed = very fast
Capacity = 50% wasted space
This type is a combination of the first two types. the first pair of disks act in RAID 0 mode providing high throughput. The second pair, also in RAID 0 mode, act as a mirror of the first providing redundancy. The number of disks in the array can be increased always in pairs to gain both speed and capacity up to the maximum drives supported by the mainboard or used disk controller.
What is RAID 5?
Minimum HDDs required = 3.
Hardware failure protection = high
Speed = very fast reading, mid to fast writting
Capacity = 1 disk in the array is lost to data parity storage. The more drives in the array the less wasted space.
In this type one disk stores redundant data and the other two store 50% of the data (3 disk array). When reading from disk it performs as a RAID 0 array providing the best performance but when writing, the OS or the controller have to perform data parity calculations reducing throughput performance. How much performance is lost? It depends on available CPU power of the PC (OS RAID) or of the hardware RAID controller's processor. Provides good protection against hardware failure (as long as only one drive fails...) but depending on drive size and RAID system (software Vs hardware, high end Vs low end) array rebuilding times can get quite lengthy. In the unusual case of two consecutive drives failing before array rebuilding is finished chances of data recovery can get quite slim.
What is RAID 6?
Minimum HDDs required = 4.
Hardware failure protection = very high
Speed = very fast reading, mid to fast writing
Capacity = 2 disk in the array are lost to data parity storage. The more drives in the array the less wasted space.
Identical to RAID 5 except that instead of one drive used for parity two are used doubling the hardware failure tolerance. This type is usually only provided by add on hardware RAID disk controllers. Is not common in home environments due to higher costs.
What are RAID 50 or RAID 60?
They are two types of RAID arrays are usually only found on big size enterprise storage systems. By combing RAID 5 arrays of RAID 0 arrays They provide extreme speeds and fault tolerance but they are not within average Joe's budget. If SAS controllers with port expanders are used a RAID 50 or 60 can comprise hundreds of HDDs. Special rack mounted cabinets are used with hot swappable trays and redundant power supplies and cooling systems.
Important options to consider when working with RAID arrays.
Hot spare disks.
Some controllers or OS RAID apps (including Intel's on-board RAID) offer the possibility of designating a disk as a "Hot Spare". These disks cannot be used for anything else, have to stay unpartitioned or formated and have to be of at least the same size as the biggest disk in the array. When a disk in an array fails the OS or the RAID controller automatically uses this "Hot Spare" to start rebuilding the broken array in the background. This is not only convenient but as it acts as soon as the failure is detected it reduces the chances of data loss due to another disk failing.
High end controllers allow for both "dedicated" and "general" hot spares disks, this means you can assign a hot spare to an specific array in case you have arrays for different purposes (IE an array of fast disks of small size and an array of large capacity but slower disks).
Hot swap disk bays.
They are cases for mounting HDDs in a way that makes it very quick and easy to remove them from the computer, replace the disk inside and put them back in the computer. The are probably only relevant on systems running 24/7 with some criticity (IE. enterprise servers).
Copy Back protection.
After a disk failure and the RAID controller has used the available "Hot Spare" disk to rebuild the array, when the user replaces the failed disk with a new one, Copy Back restores the system to the fresh drive and the "Hot Spare" disk recovers its protective status again.