I’d like to get some better performance out of our network. We currently have a gig plan with Xfinity. This morning I ran speed tests on my cell phone. The first connected to the modems wifi reported 600Mbps. The second connected to the wifi provided by the TPLink access point was 120Mbps. The 600Mbps seems perfectly reasonable. I also know that I’m going to loose a little speed going through pfsense and the switches, but I think I should be able to get more than 120Mbps.
The link between the omada switch and the aruba is just a cat6. I feel like I’ve lost perspective about what I should be doing to track my issues down. What should I look at first to try and pin down my speed issues?
Your two tests aren’t really apples to apples. The modem’s Wi-Fi skips pfSense and your switches completely, so the slowdown could be anywhere in that chain. My money’s on pfSense (especially if it’s a VM, since there’s a reason they call it the “forbidden router”) or the TP-Link.
I’d break it down like this:
Plug in. Wire a laptop or desktop to a switch and run a speed test. If you get near gig, your Wi-Fi is the problem. If you’re still around 120, look at pfSense.
Test the AP by itself. If you’ve got a spare machine, run an iperf3 server on it wired to the LAN and test from your phone on the TP-Link. That leaves pfSense out of it.
If it’s pfSense and it’s a VM, check that you’re using VirtIO or vmxnet3 NICs, look at your offload settings, and make sure the VM has enough CPU.
Check your band. Make sure your phone is on 5 GHz and not 2.4. 120 Mbps is a pretty classic 2.4 GHz number.
Also, 120 is above what a 100 Mbps link can do, so a bad cable or a port stuck at 100 is probably not the issue.
Start with the wired test. That’ll tell you which way to go.
My money is on the EAP610 being incorrectly configured. The phone may be using the 2.4ghz band on the Omada, or if it is 5ghz, it may be set to a channel width of 20mhz, poor channel selection on the Omada, etc.
I did a speed test on a hardwired computer and it tested at 287.
Pfsense is running on XCP-Ng and I’m passing a Intel X520-DA2 10Gb SFP+ directly to the VM. Disable hardware checksum offload was checked. I did uncheck it based on what I was reading but no real change in the speed test with the wired computer. As far as CPU goes I have 8 cores assigned and the dashboard indicates that it is utilizing about 50% of the CPU.
Thoughts?
I’ll dig into the EAP config after I get the wired stuff in a little better place
Looks like you have wired backhaul. Turn off Omada Mesh if enabled. The EAP610 does not have a dedicated wireless backhaul channel, thus mesh will share bandwidth with wireless clients.
If you are using the same SSID on the EAP610’s, make sure they are on different non-overlapping channels. Also check your channel width section. Don’t assume wider is always better, especially in crowded environments.
Use a decent wifi analysis app to check your environment and channel selections. https://www.netspotapp.com
I did a bunch of digging and working last night and I noticed that mtu was not set the same between the switch and the network card for PFsense. Based on the reading I had done, I tried to switch both ends to 1500. The problem with that is that the moment I change the Aruba mtu to 1500 the web interface of pfsense doesn’t work. For now I have all of the ports set to mtu = 9000. On my drive in to work this morning I remembered having issues with finding sfps that would work (3 or 4 years ago). I think I ended up with a generic pair that I could get to work between PFSense and the aruba switch assuming everything was “fine”.
I spoke with a friend at work today who deals with production networks (in the music industry) and he agreed that the sfps would be worth looking into. Here are the two I think might provide a better solution:
Could the wrong SFPs cause some of the symptoms I’ve been experiencing?
@elvisimprsntr I’m a little embarrased that I didn’t think about the wireless porttion of wifi. I’ll dig into that more tonight. Thanks for pointing all of that out.