Packet loss and online gaming - who to blame

timeshifter

Well-Known Member
Reaction score
2,177
Location
USA
My son and I both like to play Fortnite. This weekend, and once about a week or so ago, we had so really bad upstream packet loss, it would come and go.

We connect to a server called NA-East. My understanding is that Epic Games rents server space from Amazon Web Services. Figured I just figure out the IP address(es) of whatever server we're connected to and run some traces. Using netstat -b -n I think I found a few IPs. This one looks like one of them:


52.20.253.187

So I ran a traceroute to it. It looks like hop #7 is the last one for my ISP before it switches to another network.


Network_Utility_2019-01-13_23-56-56.png

If I play the game and run continuous pings to that address, the slow responses and dropped packets correspond to the lag I see in the game when the game reports upstream packet loss.
Network_Utility_2019-01-13_23-25-38.png

I play with a partner sometimes who is several states away, in a different time zone. When I'm experiencing the lag he has no issues at all.

In my mind this points the finger squarely at my Internet provider, not providing a reliable path to this server, and the failures I'm seeing are inside of their network as far as I can tell.

Am I correct in all of this? Am I totally missing the point? Something in between? Thanks! I'd love to understand this better.
 
Is the other player on mac or PC, just curious.
We're both on PC. I'm running the pings and traces on a Mac... it's not the machine I'm gaming on. The packet loss happens when only one of us is playing, either one, or both at the same time.
 
What's your network infrastructure like?

Have you tried with a direct connection to the modem?

Sent from my SM-G870W using Tapatalk
 
I think I had ruled out something local causing but maybe not. I think I ran pings to my gateway 192.168.1.1 and they didn’t fail when the events happened.

Bypassing all my network gear and having my PC direct to the modem would be a good test.

But isn’t it safe to say the problem is not with the server and lies in my ISP or me??
 
Also, my network is almost all UniFi gear, USG, Cloud Key, 16 port switch, AP LR. Have one dumb 8 port switch and an Apple Time Capsule.

Looked but couldn’t find anything on the controller that would help, but maybe I don’t know what to look for.
 
Your trace route indicates either throttling, or failure at the hand off from Road Runner to Amazon. Nothing you can do.

Also, the server node appears to be overwhelmed (pretty normal for fortnite honestly such a terrible game...)

tracert -d 107.14.17.210

Tracing route to 107.14.17.210 over a maximum of 30 hops

1 <1 ms <1 ms <1 ms 10.10.10.1
2 7 ms 7 ms 7 ms 10.48.128.1
3 8 ms 7 ms 7 ms 100.127.74.28
4 8 ms 10 ms 11 ms 100.120.100.12
5 22 ms 19 ms 18 ms 68.1.4.252
6 21 ms 18 ms 20 ms 209.18.36.50
7 22 ms 23 ms 23 ms 107.14.19.55
8 49 ms 47 ms 47 ms 66.109.6.1
9 78 ms 79 ms 79 ms 66.109.3.236
10 80 ms 79 ms 79 ms 66.109.6.40
11 82 ms 79 ms 79 ms 66.109.6.33
12 84 ms 79 ms 79 ms 66.109.6.31
13 77 ms 81 ms 77 ms 107.14.17.210

Trace complete.

If you see above, that's what my trace looks like. I didn't bother to resolve names. But you'll notice the far more subtle incremental in time on each step. Hopefully, it's just a transient issue, but it's obviously beyond your network, and even your local ISP network.

*Edit* Where did you get 107.14.17.210? Because that is not an Amazon address, it's a Charter address branded for Road Runner. Silly part is, it still indicates a healthy Internet connection so your real issues are still outside your LAN either way.
 
Last edited:
Where did you get 107.14.17.210? Because that is not an Amazon address, it's a Charter address branded for Road Runner.
I picked that because I think it's the last hop on my ISP's network. I suspect they'll say "it's not our problem, everything is working fine, must be on their end".

Ran some more tests tonight. While my son was playing we pinged two IPs:

192.168.1.1 (local router)
107.14.17.210 (last hop on our ISP)

You'll see that the sequence numbers pretty much line up as I started pinging both at the same time. When he saw lag in game we'd see drops on the outside IP. Often there were corresponding high pings locally, never drops but often higher spikes. locally.

compare_local_and_remote_IPs.png
 
Last edited:
Those 20ms response lags to the local gateway are something to worry about, that probably means you're running out of CPU in the router. I haven't seen that in ages, because I Untangle here at home, with an i3 under the hood of my router, CPU is forever! That could account for about half of the timeouts I see, but there is still some serious vacillation in your ISP's network to cause this.
 
Wondering if the local gateway high pings are related to the bad or dropped packets at the edge of the ISP network. The router is a UniFi Security Gateway, which I would hope has enough horsepower for a small home network.
 
Or maybe not. Forgot that that laptop was on Wi-Fi running those pings. My machine that is hardwired is averaging about 0.40ms to the gateway.

Ahh yes, that variance would be normal for wireless congestion. Which is why wifi for an FPS is a bad idea. Still, it's only exaggerating a moderately severe intermittent packet loss issue with the ISP.
 
After all that gnashing of teeth, somehow it's magically working fine again.

I called tech support yesterday and all the lady could do was ping my modem and tell me that it looked OK. She didn't really want to talk about traceroutes, etc. Said I should call when I was home and trying to play.

Later yesterday I had to call for a customer. The guy I was talking to sounded very savvy. So when I was done with customer issue I talked to him about my issue. He understood the situation. Mentioned that he did see T4 timeouts. I asked that he call me back in an hour but his shift was ending for his weekend. He's going to call me Sunday when he returns.

But, the issue I described here is totally gone. Don't think anyone did anything as a result of my reports. If their backbone routers were really getting choked I'm sure they knew about it.
 
T4 timeouts indicate issues on the coaxial network in your area, that can explain much. I assume you're on cable? If you are, have you looked at http://192.168.100.1 ?

That link should bring up a page in your cable modem, check your signal levels. Upstream should be between 30 and 50db, downstream between -10 and 10. The closer to the middle, the better. If you're at like 47 on the upstream, it only takes a bit of sun on the wrong bit of wire outside and over the connection goes. The worst part is, the issue will be intermittent so it's not uncommon to take months for the cable provider to locate, and repair the damaged wire or equipment.
 
Your upstream power levels are out of spec, and out of sync with downstream. Do you have any splitters between the cable modem and the feed to the residence? A two way splitter is going to subtract 5db from downstream, while it adds 5db to upstream. You need to get your upstream below 49db, until you do... upload is basically dialup, download right now is fine.

However, again this looks more like an area issue, something is out of balance. To be at a 50 up, you should be closer to a -7 or -8 down. A 4-5 down indicates a two way splittter, between your device and the source. That's not really a problem, but it could be bad.

Time to get an ISP tech out to sort it... but if you want to ensure the issue isn't in the house, move the cable modem outside and attach it directly to the feed into your residence, attach a laptop and hit 192.168.100.1 again. Get the numbers again, those will be as "good" as it gets. Properly working DOCIS 3.1 should be around 6-7db down, and 32 or so up. That way a single two way splitter takes you to 1-2db down, and 37-38db up. That's how you get enough signal to everything in your home! If those numbers aren't there, the provider isn't delivering good service to the side of the house, and they need to fix it. Make them test the signal at the tap outside the house, they'll note the issue, and escalate to engineering to re-balance the area. Cox has a 14 day response on such tickets... bring your patience.
 
Last edited:
P.S Those "uncorrectables" are packet loss... an ISP support agent should see this and schedule a tech... This is a cut and dry ISP issue now.
 
Back
Top