Architecture question for speed improvements

[nadavexpim]

Hello everyone
hope i am not repeating any known article i missed on google

My architecture is as follows
2 esxi servers with 1 nic for management and 2 nics-lag for san (ISCSI)
1 netapp storage with 3 nics-lag for san (provides storage for esxi machines)
1 VEEAM AND VCENTER Physical server with local disks in raid 6 as backup storage with 1 nic for san (connected to iscsi and shows esxi datastores physical luns from netapp - offline as described in manuals) and 1 nic which is connected to the same network as esxi management ports this server is also the proxy (it is very strong)

my bottleneck is 99% source all others are under 20% - my backup usually runs at 85mb/s
i have one free nic -
the question is
to improve performance - should i create a lag on the iscsi nic which reads the data in san mode from netapp, or should i lag the lan nic which talks with esxi on management nics?

hopefully my question and architecture is understandable enough, and maybe someone can give me an answer
thanks in advance
nadav

[dellock6]

Hi nadav,
85 MB/s is not the limit of a 1 GB link, there are still 40 MBs before reaching the limit (in theory 125 MBs, practically something less due to overheads and protocol efficiency). So, the doubt is the bottleneck is exactly the Netapp storage. You didn't told us any further info about it, which model is and how is configured?

Sounds from your description you are using directSAN access, in this case it could be that the proxy can saturate the source storage...

Luca.

[nadavexpim]

Hi luca and thanks for the reply
the netapp is fas2240, it has 3 nics as a lag, which contains a big lun of 12tb and from the deduped space i have another lun at 4tb the esxi and the veeam connectes to the storage via iscsi
as i stated the esxi hosts has 2 nics for the iscsi storage and the veeam has one nic for iscsi as read only
what more information can i proivde to help with understanding the issue?
by the way all disks in netapp are 10krpm sas in dual parity aggregate

[dellock6]

what I'm saying is, maybe is not an issue, but simply that is the maximum speed you can get from that storage, because of that 99% bottleneck on source.

Anyway, as I asked you before:
- is the veeam proxy a physical proxy?
- if yes, are you sure from backup log that DirectSAN is used? Or maybe is failing over to network mode...
- 2 nics on ESXi doesn't matter if you use directsan, the limit is anyway 125 MBs, that is the single link of the veeam proxy

Luca.

[nadavexpim]

Hi luca
the proxy is also the physical server where vcenter and veeam reside
im guessing the bottleneck really is the netapp although it seems strange because it supposed to be a really strong storage
maybe i need another proxy as a veeam on one of the esxi hosts?

[dellock6]

If the bottleneck of the job is source, an additional proxy can only place more load on it and get things worse.
If you want and you have resources, you can try to add a virtual proxy, be sure to have parallel processing enabled, and see if statistics improve.

Luca.

[nadavexpim]

i will check it out and write down the results
thanks a LOT!