7200 RPM disk enough for double deduplication?

[conradblack]

Heyo,

I'm speccing out a server to run veeam on...and I have a choice between 7200 RPM SAS disks that have more disk space and 10k RPM SAS with less disk space for our repo.

We plan on having Windows Server 2012 dedupe running on the volume where our Veeam repo will be and having dedupe turned on for our jobs. We'd prefer to take the bigger disks but will the speed of them affect the performance in this scenario?

Any insights would be appreciated!

Thanks

[Delo123]

Dedupe process hits the disks quite hard but could be interupted if needed. However nearly all reads from dedup volumes are random reads since blocks are scattered. (So it's not really a "rehydrate" thing, since cache can do magic there), but as said for random reads you need iops, so certainly go for 10K-15K SAS. If possible even have a look at semi enterprise ssd's (like Samsung OEM SSD, DC EVO or something) and maybe do some tiering in a windows storage spaces

[ACC]

In all honesty, its impossible to give any sound advice as we dont have any information about your environment and backup.
- How large is the backup?
- How many backups will run simultaneous?
- Backup window?
- Max recovery time?

On deduplication, its recommended to do it only at one point, its either Veeam, Windows or storage level.

[foggy]

On deduplication, its recommended to do it only at one point, its either Veeam, Windows or storage level.

Actually you can have both Veeam inline and Windows deduplication enabled, there's not much in terms of impact.

[conradblack]

Dedupe process hits the disks quite hard but could be interupted if needed. However nearly all reads from dedup volumes are random reads since blocks are scattered. (So it's not really a "rehydrate" thing, since cache can do magic there), but as said for random reads you need iops, so certainly go for 10K-15K SAS. If possible even have a look at semi enterprise ssd's (like Samsung OEM SSD, DC EVO or something) and maybe do some tiering in a windows storage spaces

Yeah, I figured more iops would be better for performance...but there's a serious price discrepancy between the 7200 and 10k disks. The disks I'm looking at are 2T, 7200 rpm near-line SAS so they're enterprise grade, just not 10-15k.

In all honesty, its impossible to give any sound advice as we dont have any information about your environment and backup.
- How large is the backup?
- How many backups will run simultaneous?
- Backup window?
- Max recovery time?

On deduplication, its recommended to do it only at one point, its either Veeam, Windows or storage level.

Around 2T, and it'll be one job followed by the next...each job contains 2-4 VMs. The backup window is about 10 hours, and we don't have a specific max recovery time, but obviously the faster the better.

[Delo123]

2TB is total backup size? Then it should fit your backup windows no matter what...
Easiest thing is to test an active full, that will be your max runtime (if no other IO is being done).
Should you go for 7.2 drives maybe you could consider 2 non enterprise ssds in raid 1 as a cache for storace spaces or controller.
We also use mostly 7.2K drives for backups (sata) as the will only be in the system for 6 months before we swap them.
However we have 72 drives in multiple raids striped together which gives us more then enough troughput.
Multiple vdisks/volumes are created in windows as backup repository so we can have multiple dedupe jobs running which makes it a bit easier handling big backup files

Dedupe in general in windows however is a great thing (as long as it works:))

FreeSpace SavedSpace OptimizedFiles InPolicyFiles Volume
--------- ---------- -------------- ------------- ------
40.35 TB 25.36 TB 388 388 G:
51.97 TB 21.89 TB 198 198 I:
50.79 TB 5.35 TB 9854 9854 J:
29.7 TB 5.38 TB 159 290 H:

Still hoping Veeam will someday support splitting backup files in chunks however.....