E-Series Backup Repository Volume Configuration

[chjones]

Hi all,

Hopefully a quick question.

With the NetApp E-Series (or whichever vendor it's purchased from as they're all just rebadged NetApp devices), I prefer to use Dynamic Disk Pools (DD) for the storage but wondering if there is any recommendation on how to present volumes for use as backup repository. Is it better to:

1. Present one large volume and Veeam to use it all
2. Present smaller volumes for each backup job sized appropriately for that job and it's retention requirement?

Option 1 is the easiest to manage. However, Option 2 could improve the performance as you can increase the number of simultaneous writes to the storage.

Just after some real-world feedback on how to choose the best performing design.

[nmdange]

I'd say just do two volumes so you can get some throughput on both controllers. I don't really use these units for backup repositories, but two volumes seems reasonable to me. I'd also stick with thick volumes, so you don't lose caching. Also if you do run into performance issues, the SSD cache feature might be worth considering.

[chjones]

Thanks mate, much appreciated for the reply, and makes sense. Cheers.

[dharenkamp]

Hi,

NetApp has a BP Document for Veeam on E-Series:
https://www.netapp.com/us/media/tr-4471.pdf

In short:
optimal performance
- Raid6 (8+2)
- Multiple Volumes (even number)
- Volumes thick provisioned
- 512KB segment size

large configurations
- DDP
- even number of volumes in DDP
- Volumes thick provisioned

in general:
- use Server 2016 with ReFS
- install multipathing driver from SANtricity
- for iSCSI: use different VLANs / Subnets for iSCSI multipathing

IHAC that is using such a iSCSI setup with DDP and we´re at the maximum of the 10G links.
There are 120 NL-SAS disks in place

[chjones]

Thanks Daniel. You're a legend.

I even looked at that TR and completely missed that section on NetApp E-Series Volume Configuration Guidelines.

Champion! Thanks mate.

[chjones]

Quick question. One of the stated benefits of ReFS for Veeam is the Fast Cloning capability, which is meant to help with Synthetic Fulls.

What backup method are you using? I'm personally a fan of Forward Incrementals with Weekly Active Fulls. It's what I've used for the past 6 or so years. Are you using Synthetic Fulls or Active Fulls, and how often are they being created?

[chjones]

Sorry for the bombardment of questions. I'm reading up on ReFS and will no doubt have questions about it that I need to find answers to.

You mention you are using the 10Gb host interface ports on the E-Series, so that means using iSCSI from your Backup Repository Server. Everything I can find regarding ReFS says that direct-attached storage is supported, but not SAN storage, which includes iSCSI. Have you found anything that suggests otherwise? I'm hesitant to propose an unsupported solution for my customers.

[dharenkamp]

Hi,

with ReFS you have an advantage of block cloning support.
So your Syn Fulls will be about 5 times faster.

I prefer to use syn fulls every week and 1 active full every month with traditional backup.

In this case the iSCSI interfaces are used to connect the repository Servers to the eSeries.
But you can also use SAS or FC.

And yes: official statement of MS is that ReFS on SAN is not supported.
We´ve decided to use it because it is a big environment with about 4000 VMs and with ReFS the Syn Fulls are much faster.

Here is a Link to a post where this topic is discussed for a log time.
https://forums.veeam.com/veeam-backup-r ... 46019.html

Regards
Daniel

[nmdange]

I am a big fan of using SAS-based controllers with these units. You get a lot more throughput compared to FC/iSCSI.

[dharenkamp]

I prefer also SAS when it is only one (or two) physical server.
In out case the repositories should be high available and so we´ve decided to use iSCSI and the repository servers are vitual machines on a vSphere cluster.
So the repository servers are HA protected by vSphere.

[RubinCompServ]

for iSCSI: use different VLANs / Subnets for iSCSI multipathing

What do you gain by using different VLANs for multipathing if they are still using the same 10Gb links?

[dharenkamp]

It´s always better to have everything redundant.

So if you use different VLANs you have different source and target IPs.
If there is a configuration error or anything else you only loose a path, not the Lun connection.