Please explain RAID5

[Katanand45]

I'm in the proces of setting up a NAS with a raid configuration. My first thought for data security and little downtime should there be a drive faliure was RAID1. If a drive fails the mirror takes over. I can replace the drive when I want. The end users wouldn't even know there were a problem.
The problem with RAID1 is it reduces storage space in half.

If I were to go with a RAID5 is all the data accessable if a drive fails or will I have to replace the dead drive first? What's the recovery process on RAID5?

 

[ch33zw1z]

Googled: http://en.wikipedia.org/wiki/RAID

The data will remain accessible (assuming you don't exceed the drive failure tolerance on the array, ie more than one drive...depending on the array) on the array after a drive failure, the controller knows what should be on the missing drive based on the parity bits, performance with be very much slowed while the controller spends time recreating the data when requested by a user.

The recovery process for RAID 5 is replacing the failed drive, the controller rebuilds the data on the drive from the data+parity bits on the remaining drives. During this rebuild performance will again be slowed down until completed.

 

[Fullmetal Chocobo]

Good diagram of RAID 5.

All of the data is available in a RAID 5 array if a drive fails. However, it operates in a degraded state: there is no redundancy at that point, and performance is greatly diminished due to the fact that the controller has to calculate the data from parity instead of just reading it. The rebuilding process differs by controller, but it is usually just a matter of replacing the bad disk and letting the array rebuild. All of the controllers I've used (Highpoint, Areca, etc) do rebuilds automatically.

 

[mcv]

Note that storage isn't halved in RAID-1. The point of RAID is redundancy, not increased storage space, and RAID-1 offers exactly the same storage as the smallest drive in the array (which can be bigger than 2 disks, in case you're paranoid). Other forms of RAID offer more space than a single disk, but the storage space of the array will always be less than the sum of the space on the individual disks. That's the whole point.

 

[JackMDS]

This is a good explanation to all level of RAID.

http://publib.boulder.ibm.com/...cinfo/fqy0_craid0.html

 

[Katanand45]

Originally posted by: mcv
Note that storage isn't halved in RAID-1. The point of RAID is redundancy, not increased storage space, and RAID-1 offers exactly the same storage as the smallest drive in the array (which can be bigger than 2 disks, in case you're paranoid). Other forms of RAID offer more space than a single disk, but the storage space of the array will always be less than the sum of the space on the individual disks. That's the whole point.

If you want a 1TB RAID 1 configuration you'd need 2x 1TB hard drives. This is what I meant by "half the storage"

 

[Madwand1]

Originally posted by: Katanand45
If I were to go with a RAID5 is all the data accessable if a drive fails or will I have to replace the dead drive first? What's the recovery process on RAID5?

RAID is not a standard, so all the implementations differ to some degree, and most differ significantly. In all cases, you should try to learn about your particular system's failure handling beforehand -- i.e. go through all the what if's in actual practice with your system of choice before going live with it. Learn then if it matters which order your drives are connected (it probably doesn't), if you can expand the RAID array online (depends on the implementation), etc.

And of course remember that "RAID alone is not a backup", and have something that will recover your data for all the cases that RAID doesn't cover, and for when the RAID implementation itself has a fault.

In terms of performance, RAID 5's Achilles' heel is write performance. If that matters, be sure that your system is capable of high-performance writes and configured accordingly. For high-reliability systems, this implies a write-back cache with battery back-up.