You will need to set the tunable to get multiqueue on the vmx NICs:
https://docs.netgate.com/pfsense/en/latest/hardware/tune.html#vmware-vmx-4-interfaces

I would still expect to have seen better than 320Mbps though unless the per core speed is low.

Steve

Thank you so much for that link, massive improvement just from that. I'm now seeing 600-800 Mbit/s with speedtest running on both the pfsense VM and also my ubuntu VM. That's much closer to what I'd expect to be getting.

My server is really old and it has an i7 950 @ 3.07GHz but thought it should be good enough for this purpose. The pfsense VM is currently configured with two virtual cpus; I'm not sure if giving it more CPUs would make a difference after making this change, but I can experiment with that.

I would still expect to have seen better than 320Mbps though unless the per core speed is low.

Steve

switch that my server's LAN port is connected to? If that's the case

No - you need to test actually routed through pfsense..

laptop -switch - lan (pfsense) wan -- modem -- internet

You need to test what pfsense can actually route.

Pfsense is not meant to be a client in a speed test.. its a "router" if you want to test what speed you get then route through it.. Don't use it as some client.. If you have a beefy enough pfsense box or slow enough internet, its possible you can see your full internet speed running the cli.. But its also possible you won't see what it can do via routing the packets vs trying to actually process them..

While its fine to validate your seeing traffic, or once you know what is full speed for the cli after you have validating your routing at full speed, etc..

But I would not take cli running on pfsense as same as you would see when routing.

@punchy said in Slow internet access within VM:

cli python package on both the pfsense VM and ubuntu VM results in speeds of 200-400 mbps

This while a test, really isn't a valid way to test what the pfsense vm can route/firewall - you need to route through this vm with your speedtest client to validate it can route/firewall at the speed your isp provides.

Would the following test cover what you're proposing: laptop connected with ethernet to the same switch that my server's LAN port is connected to? If that's the case, then I've done that test at I'm getting close to gigabit speeds using speedtest in the browser.

I do wonder though, what sort of issues do you see with using the speedtest cli command on one of the VMs to test my internet throughput? I'd expect that to be bottlenecked only by virtual network infrastructure, but for some reason the internet speeds are almost a third of what I'd be expecting.

cli python package on both the pfsense VM and ubuntu VM results in speeds of 200-400 mbps

This while a test, really isn't a valid way to test what the pfsense vm can route/firewall - you need to route through this vm with your speedtest client to validate it can route/firewall at the speed your isp provides.

I only have two network ports on my server, both on the motherboard. They're configured in VMWare without any pass through, the VMs are getting virtual interfaces. The modem is connected to one of the network ports, and the switch is connected to the other. The switch is what everything else in my network connects to (that is outside the VMs). This is the configuration I used during the test and daily as well. The only time I had a different configuration is when I connected my laptop directly to the modem, but otherwise it's the same test setup.

laptop connected straight to the modem using ethernet I get close to 1gbps (high 900s) when using speedtest

It seems like you should be getting more here...like 1450gbps. Never used ESXI for pfSense but I know for Proxmox/pfSense VM, it's important for WAN/LAN to passthrough. Is the switch, when testing, connected directly to pfSense LAN?