Speed Issues Veeam-VMware-Nimble

[brlomba1]

Looking over our backup job speeds I feel like these should be much higher

Setup: Vmware 6.5, Nimble array for the Virtual machines, separate nimble array for the backup repositories. Enabled backup from storage snapshot, connected to the storage via iSCSI network connections.

Job Stats: Throughput 160MB/s - 250MB/s
Veeam server: (Virtual, 6cpu 22GBram)
-NIC 1 10Gbps link - General network 1.7Gbps send/10Mbps Receive
-NIC 2 10Gbps link - iSCSI network 1.7Gbps Receive/ 10Mbps Send

Bottleneck stats from yesterday - Load: Source 98% > Proxy 47% > Network 6% > Target 1%
*Note - I added 2 cpu today (Bringing it to 6) on the veeam server and adjusted all backup infrastructure to handle 6 concurrent tasks to possibly help with the Proxy performance.

I'm wondering if there is some configuration I am missing somewhere to help with this. I can tell the Nimble arrays are not being pressured, the NIC bandwith remains below 2Gbps and CPU is decent on the veeam server, Some cores go to 100% sometimes but most are steady around 40%. I find it odd that my iSCSI network on my VM never goes above 2Gbps no matter what but the speed is reported as 10Gbps per Nimble connection manager

[mkretzer]

Whats kind of strange: I just went to the forum and wanted to open a similar thread, then i saw that there are already two speed threads open - our speed is also not what it used to be, no matter if we use NBD or hotadd. At first i thought REFS was the issue again but we have source as bottleneck for all these jobs.
The thing is: We have two completely different storage systems, one is a brand new all flash system which shows latencies way under 0.5 ms all the time, still it is not one bit faster to backup than our old SAN. This is highly unlikely as all our applications report perfect performance on this system.

Can this be an issue of Update 4a?

[foggy]

Please check if storage snapshots are being actually utilized - this information is available in the job session log.

[brlomba1]

They are. Its evident in both the logs and on the nimble its self

[foggy]

Since the source is reported as the bottleneck, I'd look at the source storage itself. Feel free to ask our engineers to review your setup more closely.