Question in regards to the Infrastructure/setup.

[TieT]

Hi,

Let me first explain the infrastructure we use:
1. Backup Server HP DL380 G8 (where VEEAM is installed)
- 12Gb ram
- 1 lun 18Tb
- 2Gb Trunk in the VM network
- 2Gb trunk in the iSCSI network

2. 2 esx servers DL360 G8
- 2 ports in the VM network
- 2 port in the iSCSI network

3. HP Storevirtual P4500
- 4 nodes in raid 5 (internal)
- Network raid 10 (nodes)

4. 1 Virtual server for Vsphere
- 1 port for network connection

When i'm doing backups to the backup server, i get a 90% bottle neck on source and a Throughput of 40MB/s.

Thats kinda low imo.
I'm trying to find where the problem is.

I can see that the proxy uses the network trunk on the backup server.
The backup proxy is set up in automatic mode and failover to network mode if primary transport modes fail.
Max concurrent tasks is set to 4.

[TieT]

Update

- I've presented my lun's to the VEEAM server (in read only mode) and configured microsoft iSCSI initiator to connect to the volumes.
Now i'm getting a Throughput of about 70-80 MB/s

But i'm still not satisfied, i want to get at least 100MB/s and i know its possible with the hardware i'm using.

[veremin]

I have noticed that you're using StoreVirtual as underlying storage. You can also try to utilize Storage Snapshots functionality and see whether it increases backup performance. Thanks.

[TieT]

Hi eremin,

Yes i know, but i don't want to have storage snapshots on the SAN.
i noticed when i select to have 10 restore points to keep on disk, it also keeps 10 storage snapshots...

[veremin]

In fact, the situation is slightly different. Once the backup process is finished, the corresponding SAN snapshot should be automatically removed.

May be you've mixed it up with Veeam Explorer for SAN Snapshots?

Thanks.

[TieT]

Nope,

When i did the upgrade from v6 to v7 it automatically switched to storage snapshots for my backup jobs, that resulted in a mail from the SAN saying it was @ 90 % of its capacity.
I looked directly in the SAN management console i found 20+ storage snapshots

[veremin]

If these snapshots were created during backup activity, they should have been deleted immediately after the backup was finished. In other words, the described behavior is unexpected. Some of the previously existed issues were fixed in recent patches. If this issue is reproducible, you'd better contact our support team and let them investigate it directly.

Thanks.

[TieT]

Ok, I'll give it another go and let you know the outcome

[TieT]

Ok when using storage snapshots i get the same Throughput.
No real performance advantage...

The snapshots get deleted automatically so that's good

But as far as the speed goes, am i forgetting something ?

[Vitaliy S.]

Hi Tim,

I wouldn't expect much difference in performance when using backup jobs from SAN snapshots. One of the main benefits of it is to offload the snapshot creation/deletion operation to the storage array.

Now i'm getting a Throughput of about 70-80 MB/s

Do you have it for full or incremental job passes?

Thanks!

[TieT]

i use it for an incremental pass, with creation of fulls on saturday.

Can i use SAN Snapshots when i've got a server with a mapped RAW lun attached to it ?

[veremin]

Can i use SAN Snapshots when i've got a server with a mapped RAW lun attached to it ?

Are you talking here about VM with RAW LUN attached to it or about something different? Thanks.

[foggy]

- I've presented my lun's to the VEEAM server (in read only mode) and configured microsoft iSCSI initiator to connect to the volumes.
Now i'm getting a Throughput of about 70-80 MB/s

But i'm still not satisfied, i want to get at least 100MB/s and i know its possible with the hardware i'm using.

Have you tried these recommendations for increasing backup performance when using direct SAN over iSCSI?

[TieT]

Yes, i've read and tested those netsh commands
No mayor improvement detected

[foggy]

Actually having 70-80 MB/s on the full job run is not that bad. Could you please elaborate on why are you sure your setup is able to provide more? Btw, what is the bottleneck for this job when it's using direct SAN access?

[TieT]

I think i can get about 100MB/s on this type of storage, if so it would greatly reduce my backup window.

Bottleneck is source 99% proxy 27% network 1% target 3%

[TieT]

Are you talking here about VM with RAW LUN attached to it or about something different? Thanks.

Yes it's a VM with a RAW Lun attached to it

[veremin]

Yes it's a VM with a RAW Lun attached to it

Please be aware, that the raw LUN connected to VM will be automatically skipped from processing during backup/replication activity. Thanks.

[Vitaliy S.]

Just a small correction - if it is a virtual RDM, then it will be backed up. Vladimir is talking about physical device mapping.

[TieT]

I get this error
3/11/2014 12:01:53 :: Processing 'ServerID' Error: Client error: Unknown element format: 'NAA:6000eb33a40ca0a40000000000000075695343534944/00000'

It's not a virtual RDM but a physical device mapping.
So the only possible backup for this machine is to use VM snapshots and not storage snapshots ?

[Vitaliy S.]

In order to backup VMs with Veeam (or any other image based solution) you should have an ability to take a snapshot from a VM. It is required no matter whether you use storage snapshots or not.

[TieT]

yes i know, i'm just looking into backup options for this particular VM (with RDM attached)

[Vitaliy S.]

Well...the way to go would be either converting it to vRDM/vmdk format or use agent-based solution.

[veremin]

Can you elaborate on the exact reasons of using pRDM, instead of vRDM/vmdk? Thanks.

[TieT]

Can you elaborate on the exact reasons of using pRDM, instead of vRDM/vmdk? Thanks.

The VM wasn't setup like we normally do (test), i'm going to change it to a vmdk

[veremin]

There is the blog post regarding pRDM/vmdk conversion; might be helpful for you. Thanks.

[TieT]

I just checked the setup of the RDM and noticed it is a Virtual Mapped raw lun, so veeam should backup it when using storage snapshots ?

My bad, i assumed it was a physical without checking it ...

[foggy]

Yes, virtual RDM LUNs are supported.

[TieT]

Yes, virtual RDM LUNs are supported.

when selecting SAN snapshots ?
because i get the error
3/11/2014 12:01:53 :: Processing 'Servername' Error: Client error: Unknown element format: 'NAA:6000eb33a40ca0a40000000000000075695343534944/00000'

[Vitaliy S.]

Tim, in this case I would suggest contacting our technical team for further logs review and troubleshooting.