Need help gigabit performance
-
-
To me VM should only for testing, once you are in production, should run on a dedicated, or "bare metal" as he says.
-
Thanks, I thought I would be able to run this box in a VM as its for home only not in production.
-
I would not expect the hypervisor overhead to make that much difference normally. It looks like you could probably tune that better. You might want to ask in the virtualisation section.
Steve
-
What Ethernet cards are you using in your VM?
Are you using Virtualised ones? vmxnet3 in Vmware or VirtIO in KVM?
Have you turned off all hardware offload?Tim
-
I had the offload checked. I was using the e1000 nics and the ethernet cards and its a C2000 SoC I354 Quad GbE Controller. From the supermicro 5014 board from http://www.supermicro.com/products/motherboard/atom/x10/a1sri-2758f.cfm
-
I have a similar setup and you should have no problem pushing that kind of traffic ….
WAN provider: Verizon FiOS (Gigabit symmetric)
System: Supermicro SYS-5018A-TN7B (A1SRM-LN7F-2758 system board)
Interfaces:
WAN - ix driver/ Port 1 on an Intel X520 (10gb) PCI3 board with 1gb SFP+ installed
LAN - ix driver/ Port 2 on the same Intel X520 (10gb) PCIe board with 1gb SFP+ installed
OPT1GUEST - igb driver/ gigabit port on the SuperMicro built-on i350-AM2 controller (guest VLAN)
Installation: Bare metal
System drives: Pair of SanDisk SSD Plus SDSSDA-120G drives; using a zpool mirror configuration.The performance is excellent. Absolutely excellent; zero concerns here.
I did add 3 system tunables for the X520 card:
kern.ipc.nmbclusters=9168192 kern.ipc.nmbjumbop=524288 hw.intr_storm_threshold=10000
But otherwise, things are running very well out of the box. See attached speedtest.

 -
-
https://forum.pfsense.org/index.php?topic=145162.0
Is the above the same problem? VM overhead?
-
No probably not. The overhead from running virtual should not be that large if the hypervisor is setup correctly. And on your hardware you shouldn't be getting even close to any limit at 180Mbps. Assuming you meant bps.
Steve