netapp direct san via fc software hba possible?

[fraghunter]

Hi!
I´m searching for information to make backups directly via a fc software hba (virtual veeam server + virtual proxys)
in direct san mode and put them onto a nas (cifs).

Is a config like this making sense?

At the moment I´m testing doing backups via network mode at 10g to a NAS but nothing higher than 40-50 MB/s seems possible...
The backups I´ve done to our other nas system in hotadd mode is making aprox. 40 MB/s via 1gig Ethernet to.

Our ESX Cluster and the 10g NAS are wired to a 10g HP pro curve switch...
The other nas is linked via 1g connection on the same switch...

Any ideas or approaches?!?

Best regards,

[PTide]

Hi,

Is a config like this making sense?

You could try that and it will work, however direct SAN is intended to avoid ESXi stack overhead when doing backups, so I don't see much sense in doing virtualized direct SAN.

At the moment I´m testing doing backups via network mode at 10g to a NAS but nothing higher than 40-50 MB/s seems possible...

Please make sure that your backup traffic goes via 10Gb interface, because 40-50 MB/s speed seems to be very slow for 10Gb. Also please tell me what are your bottlenecks?

Thank you.

[fraghunter]

Here are my bottlenecks:

02.11.2015 22:41:45 :: Busy: Source 79% > Proxy 4% > Network 33% > Target 34%
03.11.2015 09:43:07 :: Load: Source 99% > Proxy 32% > Network 3% > Target 0%
03.11.2015 11:16:11 :: Busy: Source 95% > Proxy 6% > Network 8% > Target 6%

Our vms are residing on iscsi storage dell md and fiber channel netapp storage fas 2040...


backup from netapp fc san 2 10g nas
03.11.2015 11:44:42 :: Busy: Source 99% > Proxy 12% > Network 8% > Target 0%
03.11.2015 11:34:13 :: hdd1 (50,0 GB) 44,0 GB read at 78 MB/s [CBT]
looks ok... but 78 MB/s in is not a lot or am I wrong?

I was testing 3 backup tasks at the same time saving the backup to our nas... the speed was 40-70 MB/s at each task.

[PTide]

What is your ESXi's management interface speed? Is it 1Gb or 10Gb ?

[fraghunter]

My management interface speed is 10Gb.

[PTide]

So, if your datastore is connected directly to host via high-speed link and your host is connected to proxy via high-speed link as well then the "source" bottleneck could mean that your datastore is on reads. Let 's make some tests to find out if it's your datastore or not:

Please create a separate job with a couple of VMs that reside only on FAS 2040 that is directly connected to your cluster via FC, choose network mode, run backup and post statistics here. Also please make sure that the proxy you choose operates at 10Gb (e1000 vNIC is 1Gb).

Thank you.

[fraghunter]

hi!
New Backups to 10g Nas:
9 servers each vm hdd from fc san is beeing processed at 30-50 MB/s the processing rate to the 10g NAS is 150-170 MB/s ....

03.11.2015 15:01:00 :: Busy: Source 99% > Proxy 6% > Network 1% > Target 0%
03.11.2015 14:20:18 :: Using backup proxy veeamproxy01 for disk hdd1 [nbd]
03.11.2015 14:20:18 :: Using backup proxy veeamproxy02for disk hdd2 [nbd]
03.11.2015 14:20:19 :: hdd1 (40,0 GB) 32,0 GB read at 48 MB/s [CBT]
03.11.2015 14:20:20 :: hdd2 (250,0 GB) 137,4 GB read at 54 MB/s [CBT]

veeam is showing my 10g Network Card beeing used...

[PTide]

Well, seems that it's your datastore, indeed. If you have any doubts then you might want to open support case in order to let our team take a closer look at your environment. If you do so please post your case ID in this thread.

Thank you.

[fraghunter]

[ID# 01098767] Backup Performance on Netapp FAS

[rennerstefan]

Hi!
I´m searching for information to make backups directly via a fc software hba (virtual veeam server + virtual proxys)
in direct san mode and put them onto a nas (cifs).

Is a config like this making sense?

Honestly I'm not a big fan of the idea to have a virtual proxy to do BfSS as you transfer the whole data at least two time through the hypervisor.
So you will need at least 2 interfaces (one for read, one for write) to make sure that there is no bottleneck. The design becomes very complex in this case.