Comments on: Choosing a RAID Level by Drive Count

By: Enterprise Guy

Enterprise Guy — Fri, 13 Apr 2018 14:48:38 +0000

The primary differentiator when choosing between raid 5 and raid 6 isn’t how many drives you have, but the size of those drives and the rebuild time. When you’re talking 6TB SATA drives that can take days to rebuild, you need raid 6 so that you’re not unprotected during the rebuild, and also to protect you against unrecoverable read errors in large disks. If you have small fast FC drives, raid 5 is fine. Big slow SATA needs raid 6.

By: Scott Alan Miller

Scott Alan Miller — Sun, 09 Mar 2014 23:41:11 +0000

Yes RAID 50 does what you say by reducing the failure pool, but it does it in a very inefficient way. RAID 6 reduces the risk even more than RAID 50 does while costing less. So RAID 50 is eliminated by always being inferior to RAID 6. Plus even “small” RAID 5 sets today are generally reckless, so there is no efficient means of making a good RAID 50. The nature of RAID 5 is that it never makes sense in URE sensititve (read: spinning rust”) disks so, by extension, RAID 50 is just a collection on interdependent risky arrays making it bad as well. Not “as” bad as the same array with a single RAID 5, but not better than RAID 6 or something else in the same array setup.

By: Sam

Sam — Wed, 19 Feb 2014 21:37:45 +0000

Wouldn’t raid 50 make sense by striping across several smaller raid5 pools, whereby rebuild time in event of drive failure would only effect one pool, and a limited number of drives necessary to involve in a rebuild?

By: Scott Alan Miller

Scott Alan Miller — Wed, 18 Dec 2013 11:58:57 +0000

In reply to greg schulz. This is an article about choosing RAID level specifically based on how many drives you have to work with. That is the purpose of the article. I figured the title gave that away.

By: greg schulz

greg schulz — Tue, 17 Dec 2013 19:26:57 +0000

SAM,

You seem to be basing your generic RAID configuration guidelines based on spindle/device/drive count and space, perhaps with an inferred availability aspect. However while general and generic, people tend to take these out of context and as you probably know, there is more than just space capacity as there is also performance and availability among other considerations to what RAID level to use when/where.

Using generic rules of thumb can be a starting point, however what about performance of reads vs. writes, random vs. sequential, large vs. small application usage profiles, where are those reflected in your guidelines?

Also, I know you don’t like RAID 5 and while I use it for 4 drive configs including with large TB+ sized drives, they are also backed-up and protected, however I also would be the first to tell you that RAID 5 is not good for everything.

However I also use RAID 1, RAID 6 among others in different ways depending on the need, this includes hardware/system/adapter as well as software based.

Just like RAID 10 or RAID 1 or RAID 6 or other RAID levels and permutations are not applicable for everything, it all depends on the application needs, requirements which also means more than just based on number of drives.

Btw, RAID 5 is more than a slight enhancement over RAID 4 (dedicated parity), as with RAID 5, there is no single dedicated priority drive with each drive taking turns handling the parity information, and with RAID 6, the dual parity is likewise spread across the drives in a manner where all drives help out (RAID 4 Dual Parity otoh have just that, two drives dedicated to parity)

Here’s a RAID refreseher

http://www.infostor.com/disk-arrays/raid/raid-remains-relevant-really-1.html

Of course, if you are bound by what a particular system or set of hardware or software that you may have can do, that will have an impact on what you also do…

Cheers gs

Greg Schulz @Storageio