I'm back again with a question, which I hope will be my last for awhile! My openwrt network of router and 3 APs is now successfully updated from 19.07 to 22.03.2 on all 4. They are running seemingly well and speeds of 500mbps on all.
The only issue I have now is since I setup my wireguard connection. I am based in Romania and connected to Surfshark UK sever via wireguard.
When I got updated from 19.07 to 22.03, I couldn't believe how much faster and snappier the internet was, all devices were loading things almost instantaneously, smart devices like my HomeKit setup just loaded with no 'updating' lag, etc.
This was on Romanian 'unfiltered' internet when I reconfigured after fresh install of openwrt, then when I setup the UK wireguard connection, even though everything is working perfectly as in I have UK ip and DNS, and all services/streaming etc are working as I want, and even speedtest is showing 450-500mpbs, I started to experience a slight 2 second slowdown on loading webpages, reddit, etc..
everything loads still but it's definitely slower even if speedtest technically shows a good speed. It's something I can live with if it's just the consequence of using a vpn, but I'm wondering if this is something I can tweak, or have overlooked, that would be causing this lag? I know that VPNs slow down your connection usually, but seeing as the speed seems to be good, and I only had about a 50mbps hit from 550 to 500mbps when I setup the vpn, I thought this lag wouldn't happen?
Thanks so much all for your help with the recent upgrade process and resolving issues!
Maybe bufferbloat along the VPN route? What sort of latency numbers do you get from Bufferbloat Test by Waveform ?
What you are describing is latency - it is the time that it takes from the point you want to start an operation to the time that it can actually begin.
I suspect that this is the case based on the location of your vpn. Basically, when you request a web page or other service, that typically means there must be a dns lookup. The lookup goes through the tunnel (over the internet) from your location in Romania to the vpn endpoint in the uk, then to the dns server. This must then make the trip back to you, as well. Then, with the actual ip information, the actual page/service can be requested - now we go thought another set of round trip delays for that data. Add to that the fact that many sites will have data coming from other domains, too (especially with ads and such), so each time another domain is encountered, you have two more sets of round trip delays.
so I did a bufferbloat test on my iMac and iPhone...the iMac (ethernet connection) was a C, the iPhone on wifi was a D...I've never done these tests before so I'm not sure if they are particularly bad or average? is the usual to have an A? thanks for your help
thanks so much for the info! I have noticed that on my streaming services and iMac connected via ethernet etc, even iPad on wifi, that I cannot see an lag...the only devices that seem to be having an issue is iPhone (but across all apps on it safari and reddit etc. I'm wondering if there is some settings that help/affect iPhones? It's honestly not that bad so I'm tempted to just leave it, but when the speeds are high and everything else is working well I'd love to see if I can get this lag sorted too...
Another thought I had is perhaps I need to upgrade my router/APs? I have 1 main router (Linksys WRT3200ACM) and 3 APs (Linksys WRT32X).
Appreciate your help!
D is pretty bad.
First thing I'd suggest is to try is add some Smart Queue Management. Install
luci-app-sqm on just your router. Once it's installed, do a web page refresh and you should now see a new
Network -> SQM QoS, and you can walk through the setup guide at https://openwrt.org/docs/guide-user/network/traffic-shaping/sqm
You already have some starting numbers, although I'm sort of confused by your two upload values varying so much (122 Mbps for non-VPN and 11.7 for the VPN-enabled pipe???, but the 122 is on the wireless and 11.7 on wired? that's just weird).
For step 1, grab the maximum upload and download rates from your above test results, take 85% of them and use that as a good starting point. (So, 715*0.85 = ~610 and 122*0.85 = ~105.)
On step 2, I've used just plain
cake for all three ISPs I've dealt with (all ended up with A+ final results).
Step 3, figuring out the overhead in
Link Layer Adaptation is the most mind-bending part. I've only had to configure DOCSIS cable, but it was pretty sensitive to getting the overhead values correct; not sure if it is so sensitive for other upstream types.
Edit: Hey forum software, don't eat my * asterisks!