What is RAID?
- Thread starter xaqintosh
- Start date
It's taking more than one disk and creating one, logical disk out of them.
There's many different kinds of RAID.
You can take 2 disks and make them into 1 logical disk, or you can do the same, but then mirror that logical disk to another 2 disks.
You can get 3 disks(Raid 5) and have it so that if one disk fails, all the data from that disk can be restored from the parity information on the other 2 disks.
The RAID in OS X is software RAID, meaning that splitting the data between all the drives is done in software. You can buy Hardware Raid Controlers that will look after this for you and also speed the whole thing up since it's not using your CPU.
There's many different kinds of RAID.
You can take 2 disks and make them into 1 logical disk, or you can do the same, but then mirror that logical disk to another 2 disks.
You can get 3 disks(Raid 5) and have it so that if one disk fails, all the data from that disk can be restored from the parity information on the other 2 disks.
The RAID in OS X is software RAID, meaning that splitting the data between all the drives is done in software. You can buy Hardware Raid Controlers that will look after this for you and also speed the whole thing up since it's not using your CPU.
You generally only see RAID used in servers that need high disk IO performance and redundancy.
You could use it on your desktop machine if you want to, but you would need to have 2 disks that are the same size, and same model/make for it to work optimally.
You wouldn't want to use 2 partitions on the same drive to make a RAID array because that would just be slower than using the one disk by itself.
I'm not sure how fast the software RAID is in OS X since I haven't actually tested it, but you *might* see some performance increase by using 2 or more disks.
You can get ATA RAID controlers that go in the PCI slot and are compatible with the Mac. That would be a much better solution if you really want to speed up your computer.
RAID only speeds up reads/writes to the disks, so you would really notice a speed increase in application launches and file copies. You would see an increase in application performance as well, but not as significant, because the virtual memory read/write would be faster.
You could use it on your desktop machine if you want to, but you would need to have 2 disks that are the same size, and same model/make for it to work optimally.
You wouldn't want to use 2 partitions on the same drive to make a RAID array because that would just be slower than using the one disk by itself.
I'm not sure how fast the software RAID is in OS X since I haven't actually tested it, but you *might* see some performance increase by using 2 or more disks.
You can get ATA RAID controlers that go in the PCI slot and are compatible with the Mac. That would be a much better solution if you really want to speed up your computer.
RAID only speeds up reads/writes to the disks, so you would really notice a speed increase in application launches and file copies. You would see an increase in application performance as well, but not as significant, because the virtual memory read/write would be faster.
You're welcome 
RAID is the logical inverse of a partition.
A partition is taking one physical drive and making multiple logical partitions out of it. Logical Volumes or just Volumes is the usual term for the usable spaces created.
RAID is taking multiple physical disks and arranging them into 1 Logical Volume. This can be done in different ways for different benefits. For all of this assume your drives are all af the same size in a set. Most RAID drivers require this.
I get RAID 1 and RAID 0 confused, but one is striping and the other is mirroring.
In mirroring mode you write the same data to multiple disks (usually 2) so that if one drive fails, you still have all of your data fully usable. This is roughly the same speed as just a normal drive, but it keeps your data safe.
In striping, you write some data to one drive, then to another (usually 2) in round robin fashion so that any lengthy read or write involves all drives in the RAID set equally. This makes the max throughput of the RAID set the sum of the throughputs of the drives. The latency is the same as any one of the drives.
You can put RAID 1 over 0 or vice versa. For instance, involve four drives in two groups of two so that you double the throughput, and you get redundancy of 2 copies of everything. 4 30G drives would feel like a double fast 60G drive, and any 1 drive could fail without destroying data.
Raid 4 usually uses lots of drives, but at least 3. (often 5 or more) This uses all but one drive like a striped set, making access fast. The last drive keeps parity for the corresponding bits on all of the other drives in the set. This means that 7 drives act almost as fast as the sum of 6 of those drives and any 1 drive can fail without losing data. If a data drive fails, it's data can be rebuilt by reading known data, and inferring the missing data using the parity disk. Raid 4 is never used, but it's the stepping stone to understanding RAID 5.
Raid 5 fixes a problem where RAID 4, under certain (fairly common) access situations, would act no faster than a single disk. Usually the parity disk would become a bottleneck; so RAID 5 splits the parity job between the disks depending on where the data is being stored. RAID 5 also allows any one disk to fail without loss, and gives performance roughly equal to the sum of the throughputs of all but one drive in the set, just like RAID 4 on a good day. The control for this is tedious though, and doing it in software is so laborious that it is often slower than just using a single drive. Also, it's not as fast at writing as it is at reading, unlike single drives.
What I'd like to see is RAID 5 over 5, where you use 9 drives in 3 groups of 3, with a throughput equal to twice the throughput of a drive, and capacity equal to 4 of the drives. Any 3 drives could fail (in some cases 5 could fail) and the data remains intact. See if you can figure that one out. ;-)
A partition is taking one physical drive and making multiple logical partitions out of it. Logical Volumes or just Volumes is the usual term for the usable spaces created.
RAID is taking multiple physical disks and arranging them into 1 Logical Volume. This can be done in different ways for different benefits. For all of this assume your drives are all af the same size in a set. Most RAID drivers require this.
I get RAID 1 and RAID 0 confused, but one is striping and the other is mirroring.
In mirroring mode you write the same data to multiple disks (usually 2) so that if one drive fails, you still have all of your data fully usable. This is roughly the same speed as just a normal drive, but it keeps your data safe.
In striping, you write some data to one drive, then to another (usually 2) in round robin fashion so that any lengthy read or write involves all drives in the RAID set equally. This makes the max throughput of the RAID set the sum of the throughputs of the drives. The latency is the same as any one of the drives.
You can put RAID 1 over 0 or vice versa. For instance, involve four drives in two groups of two so that you double the throughput, and you get redundancy of 2 copies of everything. 4 30G drives would feel like a double fast 60G drive, and any 1 drive could fail without destroying data.
Raid 4 usually uses lots of drives, but at least 3. (often 5 or more) This uses all but one drive like a striped set, making access fast. The last drive keeps parity for the corresponding bits on all of the other drives in the set. This means that 7 drives act almost as fast as the sum of 6 of those drives and any 1 drive can fail without losing data. If a data drive fails, it's data can be rebuilt by reading known data, and inferring the missing data using the parity disk. Raid 4 is never used, but it's the stepping stone to understanding RAID 5.
Raid 5 fixes a problem where RAID 4, under certain (fairly common) access situations, would act no faster than a single disk. Usually the parity disk would become a bottleneck; so RAID 5 splits the parity job between the disks depending on where the data is being stored. RAID 5 also allows any one disk to fail without loss, and gives performance roughly equal to the sum of the throughputs of all but one drive in the set, just like RAID 4 on a good day. The control for this is tedious though, and doing it in software is so laborious that it is often slower than just using a single drive. Also, it's not as fast at writing as it is at reading, unlike single drives.
What I'd like to see is RAID 5 over 5, where you use 9 drives in 3 groups of 3, with a throughput equal to twice the throughput of a drive, and capacity equal to 4 of the drives. Any 3 drives could fail (in some cases 5 could fail) and the data remains intact. See if you can figure that one out. ;-)
ByerlyRips
Registered
Just in case anyone was wondering, RAID stands for Reduntant Array of Independent Disks.