Napp-it 19.12b / OmniOS 151034 / ESXi 6.7u3 slow transfer speeds

[crazyj]

I'd updated my system from ESXi 5.5 to 6.7u3, and OmniOS to 151034. I've been chasing other issues, but now that most things are resolved, I was getting 100-120 MB/s transfers through Windows / SMB, but am now down to around 80 at best.

Is there anything in that mix of hypervisor / or napp-it that would have changed? I'm only running 4TB drives for the most part. No SSDs.

I made a number of network changes, but by and large the issues are all on the same VLAN / subnet.

 

[gea]

To check hardware and settings, you can test network performance via iperf (napp-it comes with server and client). Then you can check pool performance via menu Pools > Benchmark. This is a series of sequential and random filebench tests with sync enabled vs sync disabled.

From tunings after a new setup, you can:
- use always vmxnet3 vnic and increase vmxnet3 buffers
- increase tcp and nfs buffers

This can be done in menu System > Tunings

Another option is Jumbo Frames but this must be enabled an server, client and all network switches incl ESXi vswitch. For 1G networks I would not stay with default mtu 1500.

From SMB, the newer OmniOS comes with SMB3. Especially for OSX you should disable signing and encryption. You can also try to switch oplock (switch back if no effect), see menu Services > SMB > Properties.

Last tuning option is to give as much RAM as possible in ESXi, try more vcpu and reduce latency in VM settings.

 

[crazyj]

I let it run iperf for a while... now I can't log in to the napp-it interface anymore. takes too long to respond. Though I can log in to the VM. Any tips on resetting the interface somehow, or checking if iPerf is still trying to run or something?

 

[gea]

To restart the napp-it webserver from console, run "/etc/init.d/napp-it restart". To check running processes, call ps axw.

It you have made a lot of changes, reboot a former/initial BE.