ld13
Honorary Master
When I saw the primary backup solution here:
http://www.afrihost.com/site/product/dedicated_hosting
What do you feel is wrong with RAID 1?
South Africa’s biggest forum. Discuss, discover, and connect with thousands of members.
When I saw the primary backup solution here:
http://www.afrihost.com/site/product/dedicated_hosting
What do you feel is wrong with RAID 1?
What do you feel is wrong with RAID 1?
It's slightly better than RAID 5... At least with RAID1, if 1 drive fails you still have something left. With RAID5, if one drive dies you have nothing left - especially since RAID5 is only being setup with 3 drives....
Imagine the horror when that poor client finds out his precious RAID5 meant absolutely NOTHING in terms of protection... Many, many ISP's sell 3-drive RAID5 servers and think, "this is the bomb!"
Sure, but you need more than 3 drives...With Raid 5 one drive can fail and the server will still be running and you can just replace the drive to rebuild the raid.
Incorrect.
Please, enlighten me.
Sure, but you need more than 3 drives...
And, if the drives are 1TB drives, you WILL loose your data.
Anyone running Raid 5 with 1TB drives might as well run Raid 0 because i a drive failure you are going to lose your data regardless.
This is because raid 5 (and raid 6) suffer from URE's (Unrecoverable Read Errors). When a drive fails in Raid 5, there is a % change that when trying to rebuild the failed drive, a URE will occur and there will be data loss.
This URE percentage goes up very quickly and with a very steep curve when your raid size goes above 1TB. Using 1TB in Raid 5 has some ridiculous stat like there is a 70% chance you will hit a URE and have data loss if a drive fails in Raid 5. This is partly why Raid 6 exists. It resets this percentage so that the 70% mark is up near the Petabytes.
You are better off running Raid 1 with 2x 2TB drives than you are running Raid 5 with 3x 1TB drives. At least in raid 1 your data is a lot more safe in the event of a single disk failure and your write speeds will be quicker.
You're saying that a 3 disk raid 5 array can't survive two disk failures. You're absolutely right.
No, I'm saying that if you have a single disk failure in a Raid 5 volume that is above 1TB in size, there is an extremely high chance that you will have data loss.
You're implying that any disk above 1TB in size has an extremely high chance of having unrecoverable read errors?
This will explain it a bit better than I can http://www.zdnet.com/blog/storage/why-raid-5-stops-working-in-2009/162 http://forums.storagereview.com/index.php/topic/34094-is-raid-56-dead-due-to-large-drive-capacities/
I see the percentage I mention is a bit off but I can confirm that so far we have hit a URE 100% of the time during a rebuild in Raid 5. This is why we only use Raid 1 and Raid 10 now.
Besides that, I am still unsure why anyone would want to use raid 5 now days with the overhead and potential issues, its just now worth it.
OK let me paint a scenario:
Raid 1, 2x1TB disks
Disk 1 fails. You replace it.
During the mirror rebuild, disk 0 has an URE.
Surely you're just as fscked?
No. With RAID 1, DATA is mirrored on both drives. you have two identical copies(excluding timestamped updates).
If one drive fails you still have 100 percent of the data on the other drive.
A raid 1 array rebuild is different to a raid 5 array rebuild.