So I have researched, built, and experienced a few RAID5 setups.
I built a NAS recently with the intention of archiving and filesharing only over a local network. It had 3 1TB Samsung HD103UJ drives in a RAID5 array running Ubuntu Server Ed. with Webmin and mdadm for software raid. The array worked fine and samba worked for filesharing. Unfortunately the raid rebuild times took many hours and write speeds dropped dramatically, which hindered the point of a gigbit lan.
My good friend became ambitious and created a RAID5 array in his desktop setup using 4 1TB drives on a Intel controller. He archived as much data as he could from external drives and finally he consolidated everything on his array. Next day one drive in the array had a few errors and it caused the array to rebuild which took over a day until a second drive caught died and dropped from the array. Now all the data is gone and he went through the arduous process of consolidating data on separate disks instead of a raid array.
What I have learned from this forum, mine and others experiences with RAID5 setups are a couple of things.
Firstly, unless you have a hardware raid controller and the money then you will have to deal with long parity rebuild times and poor performance until then.
Secondly, If one drive is dropped from the array or also while one drive is rebuilding the chances of a second drive failing is drastically higher, leaving your data at risk.
I've come to realize that RAID5 is not meant for archiving in a passive sense. Hulu, who has multiple SAN setups in many datacenters need RAID setups with parity so that they can quickly replace a dead hard drive while the array is still being used for streaming video; noting that I'm sure they have a separate SAN backing up data. This would be archiving data in an active sense because it requires that it be available 100% of the time.
I'm sharing this because I think some people don't understand the purpose of parity RAID setups, cause at first I know I didn't.
A question I do have is what other uses for parity RAID setups are there?
I built a NAS recently with the intention of archiving and filesharing only over a local network. It had 3 1TB Samsung HD103UJ drives in a RAID5 array running Ubuntu Server Ed. with Webmin and mdadm for software raid. The array worked fine and samba worked for filesharing. Unfortunately the raid rebuild times took many hours and write speeds dropped dramatically, which hindered the point of a gigbit lan.
My good friend became ambitious and created a RAID5 array in his desktop setup using 4 1TB drives on a Intel controller. He archived as much data as he could from external drives and finally he consolidated everything on his array. Next day one drive in the array had a few errors and it caused the array to rebuild which took over a day until a second drive caught died and dropped from the array. Now all the data is gone and he went through the arduous process of consolidating data on separate disks instead of a raid array.
What I have learned from this forum, mine and others experiences with RAID5 setups are a couple of things.
Firstly, unless you have a hardware raid controller and the money then you will have to deal with long parity rebuild times and poor performance until then.
Secondly, If one drive is dropped from the array or also while one drive is rebuilding the chances of a second drive failing is drastically higher, leaving your data at risk.
I've come to realize that RAID5 is not meant for archiving in a passive sense. Hulu, who has multiple SAN setups in many datacenters need RAID setups with parity so that they can quickly replace a dead hard drive while the array is still being used for streaming video; noting that I'm sure they have a separate SAN backing up data. This would be archiving data in an active sense because it requires that it be available 100% of the time.
I'm sharing this because I think some people don't understand the purpose of parity RAID setups, cause at first I know I didn't.
A question I do have is what other uses for parity RAID setups are there?