RAID6 configuration recommendation (single vs multiple arrays)

0

We have reconfigured our Apollo 4200(p840ar controller) with RAID 5 +4 global hotspares. but as we changed the design of GlusterFS in the storage for redundancy purpose with 2copies, we are planning to reconfigure it to RAID6.

We got 28 SATA 4TB disks with 7.2Krpm. Considering both 1*26+2HS or 2*13+2GHS.

Could anyone advise the pros&cons of both design. like read/write speed, chances of failiure, rebuild procedure & approx time, IOPS, etc.

Any inputs will be greatly appreciated. Thanks.

user670317

Posted 2016-12-12T07:17:44.700

Reputation: 1

Question was closed 2016-12-14T01:55:43.390

Answers

1

Raid 6 is expensive, but is more secure. Not sure if you have the option to create raid 6 + 0 which in my opinion is better because it has more performance (because of the stripe)... anyway, any raid using + 0 has the half of capacity... so balance between your needs:

Raid 5 -> Cheaper, it protects against 1 broken hdd. You lose only 1 hdd of capacity.
Raid 6 -> Expensive, it protects against 2 broken hdd. You lose 2 hdds of capacity.
Raid 1 -> Mirror. You sacrifice the half of capacity to have more secure raid. You can lose the half of hdds.
Raid 0 -> Stripe. If you lose any disk, you lose all your raid, not recommendable for production environments. Only recommended if combined (5+0, 6+0, or 1+0).
Raid 5+0 -> The combination of raid 5 and raid 0. You sacrifice the half + 1 hdd of capacity to have more performance and it can avoid one hdd broken on each subarray.
Raid 6+0 -> The combination of raid 6 and raid 0. You sacrifice the half + 2 hdd of capacity to have more performance and it can avoid two hdd broken.
Raid 1+0 -> The combination of raid 1 and raid 0. You sacrifice the half of capacity to have more performance and it can avoid one hdd broken on each subarray.

Here is a table for comparison: Raid table

A link with some descriptions about raids. Link

OscarAkaElvis

Posted 2016-12-12T07:17:44.700

Reputation: 475

Thanks OscarAkaElvis. I haven't done 60 before. Could you tell me bit more about it comparing to RAID 6 other than read & write. Thanks. – user670317 – 2017-01-04T07:42:19.760

On Raid 6 you can have 4 disks and you can lose 2 without losing the data. On Raid 0 if you have 2 disks and you lose 1, you lose everything... but it works faster. Raid 6+0 is a "mix". You can have 4 disks as Raid 6 and another 4 disks to make Raid 0 of that 4 disks. So a Raid 6+0 using 8 disks is faster (because there are 4 disks to increase speed, the Raid 0 part) and you still can lose 2 disks without losing the data. All depends of the number of disks you have available, the capacity and the space you need. – OscarAkaElvis – 2017-01-04T08:06:11.080

Thanks again, we have configured it as 2*13 disks+2 Global HS in RAID6, total there are 28 *4 TB SATA disks. As we have replica on GlusterFS, two of our Apollo 4200 nodes will be redundant in Gluster. Also available storage space taken into consideration. – user670317 – 2017-01-04T09:37:17.543

With 28 disks (4tb each one). You can do a 26 disks Raid 6+0 leaving 2 Hot Spares. With that 26 disks you have 26 / 2 - 2 = 11 . The division is because of Raid 0. And the -2 is because the Raid 6. Son 11 disks availables for space (26 * 4TB = 104TB). Not sure what is this for... but I think 104 TB is a good capacity and you' have better performance because of Raid 0 and more security because of Raid 6. If you want more security, reduce a little bit the raid and put more Hot Spares... – OscarAkaElvis – 2017-01-04T17:38:33.153

Asked: 2016-12-12T07:17:44.700

Viewed: 213 times

Active: 2016-12-12T07:54:53.287