Bizarre WiFi Routing Troubles

[ohthat]

I have no idea left to test. This leaves me dumbfounded.

So, 2 actual network connections, 1 cable, 1 DSL.

DSL is a modem/router combo, stable low ~35ms ping, different, but still stable, depending on test server used.

Cable modem, ethernet directly to computer, excellent, <30ms, stable ping, basically all USA servers of major services.

Speed test was ~91mbps. Since connection is supposed to be ~200mbps, I figured my cheap chinese ethernet adaptor was probably limiting the ethernet to 100 instead of 1000. I was right. In some early test, I used a 5G router with gigabit ethernet to pull ~140mbps, and good ping, but jitter was really high, like around 150ms in most tests, never under 100ms.

So, for the really inexplicable part.

We're now only considering the cable connection, because it's better ping, more stable ping, and like 8x the bandwidth.

I have 2 routers at my disposal, a $hit-tier LINKSYS EA3500, which has 5G and gigabit ethernet capabilities, and a trusty ol' ASUS RT-N12 flashed with a tomato firmware. This has neither gigabit nor 5G capabilities, though it has a lot of other powers, and seems pretty reliable and stable.

Now, through either router, even daisy-chained back through both routers, I get rock-solid pings, 25-38ms depending on the service, but very stable, and 91mbps throughput.

However, NO MATTER WHICH ROUTER, and I've gone through a phenomenal amount of settings changes trying to get good results, if I ping, say, Google through Wi-Fi (5g or 2.4g), I'll get pings mostly stable in the 32ms range, then outrageous spikes of 100, 250, 500, 1000ms fairly randomly, but fairly consistently every roughly 15 seconds. It's not exactly 15 seconds, maybe 12, maybe 20, maybe 30 seconds or a minute go by between the lag spikes, but so far, nothing I do cures them.

Is this the nature of wi-fi? Is it because my laptop is close to the routers? I also get ridiculously low throughput in some configurations, but I'm trying everything: disabling WMM, Auto channel, picking the best channels, going for N-only or whatever the highest form of 2.4ghz is -only, or Mixed, auto channel width, max channel width manually, etc. I'm making these changes one at a time, to try and isolate steps that improve ping stability, or increase bandwidth. And it's a dumpster fire. Nothing seems to clearly help or hurt, and no settings I've guessed with have resulted in pings that don't spike well over 150ms for at least 3s, every roughly 12s. Yet, this is not a feature of the connection itself, because plugging the modem directly into the laptop got me 91mbps with rock-solid, stable low pings for days.

Are there any guides to dealing with the entropy of wi-fi settings? I know the Ea3500 is a POS, but the RT-N12 is ol' reliable (or was until tonight). I'm gonna try to completely reset the Ea3500 again, aand make minimal changes, in different orders, and see if that can banish the lag spikes. I just want to experience as much of the joy of the phat bandwidth and lower ping of my new cable internet connection over wi-fi as possible.

Edit: I'm also going to test hooking the routers to the DSL connection, and see if the lag spikes exist in their wi-fi when the underlying connection is different. Out of curiosity. Depending on the results, this also opens even more madenning possibilities for inconsistency.

 

[thiggins]

Run your ping test from a Wi-Fi client to an Ethernet-connected LAN device to rule out problems with your WAN connection.

Did you try the test with another WiFi device? Did you try shutting off ALL other Wi-Fi devices when you run your test?

It's possible the problem is on the DEVICE end, not the router. Devices periodically scan all channels to look for possible roam targets. Some devices may do this more aggressively than others.

Check the WiFi settings for the device you are using and see if there are any roaming or band preference settings. Limiting to one band should reduce scan time. Setting roaming settings to be less aggressive may also help.

 

[ohthat]

Run your ping test from a Wi-Fi client to an Ethernet-connected LAN device to rule out problems with your WAN connection.

So WAN connection is fine. Stable pings and high throughput over ethernet through either router, both routers daisy-chained, or just straight to the modem prove the WAN link is good.

Maybe I'm not understanding exactly what you meant here, though? WAN means wide-area-network, which in this case refers to the ISP's network and the internet, as opposed to LAN, local-area-network, the devices on this side of the modem.

Did you try the test with another WiFi device? Did you try shutting off ALL other Wi-Fi devices when you run your test?

This is a good idea. Other wi-fi was off, but there could be interference, because there are lots of devices everywhere. I live in a dense apartment.

Crazy thing that doesn't necessarily rule out these possibilities: late last night, after hard-coding the channels to the same channels that Auto selected, I got a stable ping on that same laptop! It persisted overnight, no idea why. But haven't had the lag spikes every 12 seconds. We're about 14 hours in with stable goodness, and haven't observed the lag spikes on either 2.4ghz or 5ghz.

It's possible the problem is on the DEVICE end, not the router. Devices periodically scan all channels to look for possible roam targets. Some devices may do this more aggressively than others.

It's still very much possible, but unlikely. This device functioned just fine on dozens of other 5ghz and 2.4ghz wi-fi networks, with varying and similar levels of interference. Heck, it's even got a great stable ping on the DSL modem/router in the exact same environment with the exact same interferences.

Check the WiFi settings for the device you are using and see if there are any roaming or band preference settings. Limiting to one band should reduce scan time. Setting roaming settings to be less aggressive may also help.

I don't know about this. It's default options, and I want to leave my devices on defaults for these things. However, if there's any device that could be modified, it'd be the laptop, so I'm opening a million-tab google corps of discovery mission right now.

 

[thiggins]

Yes, WAN connection is your ISP connection.

Since you are in a dense environment it's likely you are suffering from competition for the same channel from neighboring networks. This would explain the intermittent nature of the problem. And, yes, this is the nature of Wi-Fi.

If that's the case, there isn't much you can do about it other than to try to find a less active channel. Don't rely on router auto-selection. Use a Wi-Fi scanner app to see what you're dealing with for number of networks and the channels they're using. You'll likely see a mess, but at least you'll see what you are dealing with.

 

[ohthat]

So, while I can't rule out the competition for the same channel causing it, I really doubt it'd do that on both 2.4ghz and 5ghz, on different routers, with totally different settings, but then stop doing it on both 2.4ghz and 5ghz. And, at the same time, against the same interferences, never do it with the DSL modem/router on either 2.4ghz or 5ghz. Of course, if I leave the settings how they are, and the lag spikes every 12 seconds come back in the late evening, I'll be pretty sure this is indeed the cause.

I was using a wi-fi scanner app, but strangely in both 2.4ghz and 5ghz the auto channel found a channel that gets better SnR than the obviously best channel based on the wi-fi scanner app. The obviously-best 5ghz channel would be, say, -55dBm, sharing with as few competitive networks on the same channel, and then the auto-selected channel would be screaming over multiple strong competitor networks, but would get to -42dBm, on literally the worst possible channel based on crowdedness. On the 2.4ghz channel, the obviously-best (based on wi-fi scanner app) got a decent -45dBm, then the autoselect channel, sitting on top of the loudest networks possible, on what I would estimate as the 2nd worst possible channel, gets an absurdly-good -28 to -39dBm. It defies explanation. I have no idea if it mattered that I hard-selected these autoselected networks, but coincidentally, after I made that change, the ping stopped spiking every 12 seconds, on both 2.4ghz and 5ghz.

Reminds me of the good ol' days, when my friends and i would put little tiki gods on our wired routers and switches at LAN parties and at home, to ward off the evil spirits. We were all networking-savvy nerds, but whoa be unto the nerd who forgot to place a tiki god on his networking device. It would cease to function properly within a few hours.

 

[thiggins]

Wi-Fi scanners don't show activity or channel congestion, which has the largest effect on available bandwidth, which is what you are after. I doubt the SNR number shown by these apps is a good proxy.

The way to get to the bottom of this is to packet capture.
https://www.smallnetbuilder.com/wir...ld-a-wi-fi-performance-analyzer-for-75-part-1
https://www.smallnetbuilder.com/wir...ld-a-wi-fi-performance-analyzer-for-75-part-2

That way, you can see what's really going on in the channel.

 

[ohthat]

So you don't think WiFi Analyzer, the app in part 2, is good for identifying the relative loudness of APs, and choosing the least congested channel? It's clearly not as good as whatever autoselect did in my case, but I thought it was generally robust.

 

[thiggins]

So you don't think WiFi Analyzer, the app in part 2, is good for identifying the relative loudness of APs, and choosing the least congested channel? It's clearly not as good as whatever autoselect did in my case, but I thought it was generally robust.

The only thing those apps can tell you is how many APs are on a channel and what their signal levels are. They tell you nothing about what those APs are doing. You don't know if an AP is idle, has 8 users all streaming YouTube for hours or one little ol' lady checking email.

 

[ohthat]

Still stable ping and excellent bandwidth, heading into the late evening hours.

Part of me wants to test things further, like different channels, see if I can replicate and isolate the cause of the jitter, but I'm scared to lose the excellent stability I've got!