Maximum number of ISCSI LUNs in an Scaleout Repository

[WinstonWolf]

Hello ,

What is the maximum number of ISCSI LUNs ( Windows Disks ) you can have in an Veeam 9 Scaleout Repository .

Background :

We plan to reconfigure our Netapp FAS2554 ( Backup Storage )
All together we have 130 TB for one Aggregat .
One ISCSI LUN in Netapp for Windows can be 16TB big .

Our Goal is to configure 8 16 TB ISCSI LUNs and create with this LUNs one big "Scaleout Repository" .
This Scaleout Repository then we mapped to our 4 Backup Jobs .

Thanks
Michael

[PTide]

Hi,

There is no limit on how many LUNs you can have in a repository, however there is a limit on a number of extents - if you use Enterprise Edition of VBR, you can create 1 scale-out backup repository with 3 extents for this scale-out backup repository. Enterprise Plus Edition has no limitations on the number of scale-out backup repositories or extents. Since you have a single physical storage there is no need to setup a Scale-Out repository. You might want to create a single spanned volume in order to be able to use the whole FAS as a single repo.

Thanks

[pkelly_sts]

This is way beyond our scale but out of interest, if using scaleout isn't there the opportunity to keep fulls & incrementals on different extents within it thus potentially increasing performance, which you'd otherwise lose if configuring as a single repo with spanned volume?

[WinstonWolf]

What you mean with spanned Volume .
You mean that you create in Windows one Volume with all of the ISCSI Disks with the Windows Storage Manager ?
But on the Netapp Side there is no way to make an Stripe over the ISCSI Volumes . And the biggest ISCSI LUN on Netapp is a 16TB fo Windows .

[dellock6]

Well, you may use spanned volumes if for example you don't have enterprise licenses so no option to use SOBR, or you have only Enterprise and Not Ent+ so the limit is for 3 extents to be part of SOBR. In the described scenario we have 8 volumes, so out of the limit of enterprise edition.

About the performance policy, it increases performance as the name says if the two extents (fulls and incrementals) are stored into different storage volumes (different disk pools or completely different machines) so you can consume the availabile IOPS of each device. Carving out multiple volumes from the same storage pool helps in cases like this where there's a limit on the size of the volume that can be created, but two extents running over the same volume will have to share the available performance, thus not creating a great performance boost.

Also, remember the performance policy adds performance at the expense of a larger failure domain, as you have now files spread over two extents instead of one. Losing one of the extent makes the other useless, while on data locality failure domain is reduced to one extent. Data locality may be used to spread backups over multiple extents where (like in this case) the need is more to consume evenly the available repositories, rather than increasing performance.

Luca

[PTide]

What you mean with spanned Volume .
You mean that you create in Windows one Volume with all of the ISCSI Disks with the Windows Storage Manager ?
But on the Netapp Side there is no way to make an Stripe over the ISCSI Volumes . And the biggest ISCSI LUN on Netapp is a 16TB fo Windows .

Yes, I mean that you can create 8 x 16Tb NetApp LUNs, connect them via iSCSI so they appear as disks in Windows, and create a single spanned volume over them. Another option is to use Storage Spaces. As far as I understand your case you just want to consolidate those LUNs in a single space to use it as a repo. If so then there is no need to use Scale Out Backup Repository.

Thank you.

[WinstonWolf]

Wow ,

Now im a little bit confused . My english is not so good . Can you explain it please in simple words what i have to do for best Performance .

I have 4 Backup to Disk Jobs .
I have an 10Gbit Network between the Productiv Storage , the Backup Server and the Backup Storage.

I have an Netapp FAS 2554 , in this Netapp i have only one Controller . I have 48 x 4TB Disks . With this and three Spare Disks i can Build an Aggregate with 8 x 16 TB ISCSI LUNS which are 8 x 16 TB ISCSI Harddisks in the Windows 2012R2 Server.
We have an Enterprise Plus License of Veeam B&R 9.

[PTide]

As Luca has explained, Scale out backup repository with performance will give you a better performance only during transforms, and only in case extents are located on different storages. For the best performance you can you can use Windows tools to create a striped volume, that will provide you with some kind of an equivalent of RAID-0. Before sticking to that approach make sure that you have a secondary backup destination configured, because as with RAID-0 a loss of a single volume will cause the entire backup set to be lost. 48 disks sounds very fault-untolerant, do you have any level of redundancy on your NetApp?

Thanks

[WinstonWolf]

Hello ,
Thanks for the explanation in short words . That means a Scaleout Repository over my 8 x 16TB ISCSI Drives in Windows makes no sense because it brings not better Performance ? Is this correct ?
I have only one Storage , then the best way as i understand your Post, for Performance is, to create an RAID0 Stripe Set with the 8 Drives in Windows .

In our Natap FAS2554 we have 3 Hotspare Disks , additionally we Backup every Saturday an virtual Full on Tape and every WeekDay the incremental Files from all VMs on Tape.
And we have also from every VM on the Productiv Storage, Storage Snapshots , 2 Weeks , 3 Days , last 8 Hours. The Productiv Storage is an Metro Cluster.

[PTide]

That means a Scaleout Repository over my 8 x 16TB ISCSI Drives in Windows makes no sense because it brings not better Performance ? Is this correct ?

Correct, although it might improve transform speed.

I have only one Storage , then the best way as i understand your Post, for Performance is, to create an RAID0 Stripe Set with the 8 Drives in Windows.

Yes, theoretically, but I would not recommend to use RAID-0 to store backup data because it's unreliable. Also there is another big question - what level of RAID do you have on the NetApp?

In our Natap FAS2554 we have 3 Hotspare Disks , additionally we Backup every Saturday an virtual Full on Tape and every WeekDay the incremental Files from all VMs on Tape.

Spare drives come into play when one of your primary drives fails. What kind of RAID do you have on the NetAPP?

Thanks

[dellock6]

I'm pretty sure NetApp is using RAID-DP as in many of their arrays.
to go back on track a bit, my suggestion would depends first of all if you have already a license allowing for Scale Out Backup Repository. If so, I'd go for it anyway, if not for increased performance, at least for ease of management of the free space, as with 8 equal extents the algorithm will place different VM backups over all the volumes evenly. Stay away honestly from RAID-0, at most you can use spanned volume.

[WinstonWolf]

Sorry what means SOBR ?
What is than the best way to use my Storage with Veeam ?

[PTide]

That's an unofficial acronym for Scale Out Backup Repository.

[WinstonWolf]

Thanks We have already an Enterprise Plus License of Veeam .
Tha means the best Solution for me is to create a Scale Out Repository from my 8 x 16 TB Disks ?

[dellock6]

In my opinion, yes. first create 8 regular repositories using the 8 luns of the netapp storage. We didn't talked about which servers are going to map those volumes, a decent windows machine can manage those volumes all together if it has enough cpu and ram, or you may want to use two different servers.
Then, group the 8 extents into one scale-out repository and use the data locality policy.

Luca

[WinstonWolf]

Thank you .
We use an HP DL380 G7 with 2 x E5645 Intel Processors ( 24 Logical Processors ) and 144 GB RAM and an HP 10Gbit Network Adapter connected to the Netapp FAS .
I think this Server has enough Power

[WinstonWolf]

But one Question, why you choose the "Data Locality Policy" and not the "Performance Policy" ?
The other Reason for the "Peformance Policy" will be the Place for our File Server Backup which can be in the Future Bigger than an 16TB of one single Repository from the Scale Out Repository .

[dellock6]

hi.
First, the server should be good for managing that storage, it all depends on the number of concurrent jobs you will run, but in general it has a lot of horse-power.

About the policy, the main reason as explained is that the underlying array is always the same, so it doesn't make any different to split backup files of a chain over multiple volumes, as the overall I/O will always be the same. You are only increasing the failure domain to two volumes with no performance benefit.

In regards to the size of a vm, again the policy doesn't affect this, in order to store a VM in a volume, when using per-vm chains the single vbk file has to be smaller than the volume. Counting an average 50% reduction, the VM can be as big as around 30 TB. But even with performance a single vbk is NOT splitted among volumes.