Teacher
•
11 Messages
Increasingly terrible latency over the past 1-2 months
I've had no problems since getting Uverse installed March of '17 until now. Beginning in mid December (somewhere around the 12-14th by my recollection), I started experiencing latency spikes in online multiplayer games. It was just an annoyance at first, small spikes of a couple hundred ms every half hour or so. Without going into all of the details in between, as it stands today I get spikes of several seconds (not ms) every 15-30 seconds. This makes online multiplayer gaming a literal impossibility. I'm just going to lay out all of the information that I have gathered so far, there's quite a lot.
First, this is on fiber, not DSL. The model/router is a Pace 5268AC. Yes, I'm on a wireless connection, no a wired connection is not a possibility, however I've tested with wired connections and the problems are the same. It is not a problem with wireless signal strength or quality. More on that in the screenshots below. I've tested this on five separate devices in from my desktop, to my laptop, to phones and tablets. All show identical results with high latency spikes and packet loss in the same places.
Here come a number of screenshots of various diagnostics I've performed.
inSSIDer capture showing there is zero interference or traffic from other networks in the area. I also have a very strong signal to my modem.
Statistics from the modem captured just today. Still perfectly acceptible signals to the modem. This is after a recent reboot of the modem. Uptime: 2 days, 8 hours. Please note, the increase in noise is because the modem reset me to "default" channels in the 5GHz band and I was overlapping the other two networks near me. I've fixed that since this screenshow and am currently showing -40db again.
More modem stats. Taken at the same time as previous.
WinMTR screenshot taken just now. The second hop is the problem spot for high latency, hops 4 and 5 always show high packet loss. I don't know if it's purposeful dropping of ICMP packets due to high traffic or actual loss.
Just a quick WinMTR with hostnames resolved.
Finally a Wireshark capture filtering only Duplicate ACKs and TCP Retransmissions on the route shown in the WinMTRs. You can see an RTO of over 2 seconds which is common for most of these Retransmissions.
A final thing to note, I'm in Cleveland. We had a very warm and mild November/first couple weeks of December. The problems started just about the same time "winter" finally hit us in mid December and the temps dropped and so did the snow. This could be purely coincidental, but it's definitely worth mentioning. I'll gladly provide any additional information needed to fix this.
EDIT: I forgot to mention, my bandwidth seems unaffected for the most part. Speedtests using various services (speedtest.com, fast.com and speedof.me for example).
EDIT2: I also realize that I didn't mention that this problem exists 24/7. There is no specific time of day where it is worse or better.
ATTHelpForums
Community Support
•
2.7K Messages
6 years ago
Hi @Darxide23,
I apologize about the issues you are having with your service. Thank you for providing all that information. It is always hard to identify latency issues. For testing purposes, have you tried connecting through a VPN to see if you are running into similar issues?
David, AT&T Community Specialist
0
Darxide23
Teacher
•
11 Messages
6 years ago
I hadn't even thought of that, but here's the results.
Looks a lot more normal through my VPN service. It is a paid VPN, by the way. I don't usually consider it for gaming because it's inherently slower than a healthy connection through any ISP.
1 Attachment
Fer7dk5
0
0
Darxide23
Teacher
•
11 Messages
6 years ago
This is still an issue. I have a bandwidth limit on my VPN, so I can't keep wasting it on this. If I have to upgrade my VPN plan because of this I guess I'll send AT&T the bill?
0
Darxide23
Teacher
•
11 Messages
6 years ago
Bumping. Still an issue.
0
0
bloomin
Scholar
•
290 Messages
6 years ago
@Darxide23
I have DSL and have the same problem, seems to be peering issues with AT&T partners, obviously they don't care about customer satisfaction since they haven't answered you. Sorry to say.
0
0
Darxide23
Teacher
•
11 Messages
6 years ago
Bumping. Still an issue.
0
Darxide23
Teacher
•
11 Messages
6 years ago
Bumping. Still an issue.
0
0
JefferMC
ACE - Expert
•
35.2K Messages
6 years ago
Looking at your very first WinMTR:
I want to mention two things, one of which you mentioned (hops 4/5 high loss) and one you didn't (W/C latency on hop 2).
Hop 4/5 loss: Most AT&T Residential internet customers (U-verse VDSL2 & Fiber) will tell you that this is normal. These just drop a large number of TTL-expired ICMP packets without spending time in the main CPU generating the error response. By the fact that nodes further down the line respond more normally, you can infer that they pass ICMP packets that aren't TTL-expired.
Worst Case latency on hop 2. This is your first hop router inside AT&T. It's interesting because it has to be passing your ICMP traffic faster than it's generating an error reply (which is expected). On copper, this isn't typically noticable. I'm not sure about fiber.
Any chance you could get a winMTR report more focused on the period when you have a latency event? An alternative is a bunch of PING -t on a variety of nodes on the path towards your server, check all of them to see which ones show something interesting at the time of increased latency (dropped packet, long latency). Might let us draw a map to where it's occurring.
1 Attachment
Capture-2018-02-05.PNG
0
0
Darxide23
Teacher
•
11 Messages
6 years ago
As I did mentioned, there are no "latency events". This is a 24/7 latency event. Any given moment, my WinMTR looks just like the one I posted originally.
Edit: Just for emphasis, I took this right now after I posted. I didn't have to wait for an "event". I wasn't doing anything where I could notice any latency. (I was just reading some blogs, not gaming).
Something you also mentioned about pinging the nodes individually, if I do a ping to the second hop (the one where the high latency spikes are appearing which I did mention as the main focus of my post) then there are no latency issues to that node when pinging individually. So I ran WinMTR on that node for a while. No problems. Then I ran WinMTR to the third node and the second node then showed it's horrible WC latency with lots of spikes. I don't know if that means anything. See image below to see what I mean.
2 Attachments
06sVP3A
4sqLGZ6
0
0
JefferMC
ACE - Expert
•
35.2K Messages
6 years ago
Two things:
First image: Please indicate what your issue here is. Your average latency is 40 ms to your target server. You appear to be in Ohio, the game server appears to be in New Jersey. What latency do you expect? If you want to say "packet loss," notice that the packet loss appears to be in the New York Telia network. You could complain to Telia... if they'll listen.
Second image: There is a difference in processing between being the destination of a PING and being the second hop on a PING when the original TTL of the packet was 2. In the first case, the router creates the standard reply. In the second case, the router has to generate the exception reply for expired TTL. ICMP Exception reports are deprioritized in AT&T residential edge routers. This "terrible" response means nothing, because it passed 100% of the packets destined for hop 3 with a TTL=3 and did so such that the latency was 3 ms, a great job. So, as long as you're not asking hop 2 to report errors, it does a fine job.
0
0