New repo, spec check pls

[lando_uk]

So I'm in need of some new repositories. The last ones we bought were R720xd with 12 x 4TB SATA RAID6 +HS

These are too slow, and the SATA disks don't seems to last very long. So I was thinking of R730xd with 16 x 6TB NLSAS RAID10, but do I use Perc HW Raid10 or Mirrored Storage Spaces ?

Over the years of the R720xd's, they have had a few hidden corruptions due to parity errors, I think its the combo of SATA and RAID6, as we have CCTV servers using the same RAID6 and they have been rock solid using NLSAS.

Any advice?

[PTide]

Hi,

So I was thinking of R730xd with 16 x 6TB NLSAS RAID10, but do I use Perc HW Raid10 or Mirrored Storage Spaces ?

I'd stick with HW RAID 10 since it is OS independent. Also Storage Spaces relies on the host's CPU, while HW RAID has its own module to perform all required calculations which means that the repository will have more CPU power for its daily routine.

Thanks

[lando_uk]

Do you know anything about the latest large NLSAS standard, like 512e or 4k drives, as Dell sell both versions same size, same price. We've been buying 4Tb for years and they seem fine, but the 6TB + models all use different tech and i'm unsure which is best for backup type workloads.

Thanks

[PTide]

Do you know anything about the latest large NLSAS standard

Not quite much, except for NLSAS being just SAS drive with 7200 RPM.

<...>i'm unsure which is best for backup type workloads.

A particular NLSAS drive may or may not be as reliable as any other SAS/SATA/NLSAS drive. You need to check specs to know for sure. Also may I ask you what's your backup policy and bottleneck statistics?

Thanks

[nmdange]

You might consider doing RAID 50 with 14 drives (2 raid 0 on top of 2 7disk raid 5s). You can get better performance than RAID 6 but you don't lose as much space as RAID 10.

[PTide]

With RAID 50 you can lose only two drives in the best case (one from each RAID 5) + rebuild of RAID 5 means hard work for the array and if another drive crashes during the rebuild then your backups are toasted.

RAID 10 allows you to lose 1 drive in every RAID 1 group(best case). Worst case is equal to RAID 50 - two failed drives in the same group will destroy your backups.

Thanks

[primo42]

Hello,

currently, our Veeam Backup Server is connected to a QNAP TS-869U-RP https://www.qnap.com/en-us/product/mode ... 30&event=2 with a RAID5 storage pool over 1 GbE iSCSI connection.
Because of performance and capacity reasons, we are thinking about replacing the QNAP with the following DAS:

- Intel JBOD2312S3SP (12x 3,5'', SAS 12Gb/s) with redundant PSU
- Intel RS3MC044 SAS 12Gb/s RAID card (LSI 3108)
- SFF-8088 to SFF-8644 cable
- HDDs: we are deciding between keeping the existing 8 4TB WD Red 5400rpm SATA and buying 4 additional or buying 7 new Seagate Enterprise Capacity 3.5 HDD 4Kn 10TB, SATA 6Gb/s.

We are planning to use RAID6.

What is your opinion on the configuration? Do you have any experience with Intel JBOD systems? I guess also the existing WD Red drives should work with the mentioned RAID controller? If we decide for the new hard drives, should we get the 4K version or the 512e?

Thanks a lot!

Best Regards,
Primož

[primo42]

Since I cannot edit my post, a note here -> Seagate Enterprise Capacity 3.5 HDD 4Kn 10TB, SATA 6Gb/s = SAS 12Gb/s .

[orb]

You might consider doing RAID 50 with 14 drives (2 raid 0 on top of 2 7disk raid 5s). You can get better performance than RAID 6 but you don't lose as much space as RAID 10.

Why it is a good idea, the problem is RAID50 is not considered as reliable with spindles over 1TB. While technically it works, the rate of loosing data with hardware failure is higher due to data density on disk. Well, you can do worst, the best I have seen is RAID5 of 14+1

RAID6 recommendation back in the days was 14+2 disks max, but spindles size was around 512GB each ! Today I I won't recommends to go over 6+2 for spindle of 4TB and over. Why ? Rebuild time ! We did a test and added 2 disks to a R6 of 6+2 (7200RPM NLSAS). It took nearly a week to compute new parities !.
RAID6 will work nicely as long as you stick to forward incremental + active

As final note, I know here we are talking about probabilities in case of failure and I don't hold the truth. Experiences sharing or horror stories are welcome !

Cheers

[nmdange]

With RAID50, the number of disks involved in the rebuild is lower. If you have RAID50 with 14drives as 7x2, only 7 drives are involved in the rebuild, compared with a 14-disk RAID6 where all 14 drives are involved, so the rebuild will be faster. For added safety, you could do 5x3 or 3x5 RAID50 with 15 drives.

Rebuild does take a long time, but I've been running a number of backup repositories with 21x2TB and 21x4TB as 7x3 RAID50 for awhile (2TB for 6 years, 4TB for 3 years) and haven't had issues with data loss. I do also have multiple hotspares. However, I also have a second server in an offsite location with the data copied with Backup Copy jobs, so even if I did lose multiple drives in the same RAID before the rebuild finished, I still have another copy

[ACC]

What is your opinion on the configuration? Do you have any experience with Intel JBOD systems? I guess also the existing WD Red drives should work with the mentioned RAID controller? If we decide for the new hard drives, should we get the 4K version or the 512e?

We cant give any recommendations unless you provide some information on your backup environment and the current bottlenecks in it.

Before buying anything you should check compatibility lists to make sure all components are certified to be compatible or buy a DAS as complete solution, this is usually more expensive but avoids weird problems due to compatibility issues. When using low rpm disks try to keep the IO workload sequential as much as possible, especially avoid reverse incremental backups as those generate a high random IO workload. This advice also applies to raid 5 and 6 configurations as they arent very capable to handle a heavy random write IO workload.

[alex1002]

Why not raid60? The Hitachi or similar SATA drives are good these days. I had Hitachi running in raid6 for 5 years zero issues.

Sent from my Nexus 6P using Tapatalk

[PTide]

Why not raid60? The Hitachi or similar SATA drives are good these days. I had Hitachi running in raid6 for 5 years zero issues.

The overall reliability, performance, and useful capacity of a RAID array depend not only on what HDDs you use but also on the amount of those. For example RAID 60 with 8 or less drives is not a good option because its capacity is the same as in RAID 10 (50% of total capacity) but it is also slower (lesser write penalty on RAID 10), rebuild longer which increases a chance of failure and so on.

[lando_uk]

Since I cannot edit my post, a note here -> Seagate Enterprise Capacity 3.5 HDD 4Kn 10TB, SATA 6Gb/s = SAS 12Gb/s .

From experience I wouldn't use SATA disks if there's the option of NL-SAS. We have quite a few large SATA and NL-SAS arrays and the later are much more resilient. Twice now the SATA RAID6 arrays have experienced parity corruptions that repair ok but leave large VBK files corrupt, needed to be deleted and started again. I'm not very confident with the SATA when using synthetic fulls.