AT&T U-verse Lag Triggers and Times - Technical Data

[Docfiord_Fowling]

Let's try and compile as much technical data on the issue as possible to help both our ISPs and SE figure out what is happening here.

I'm working with AT&T on this at the moment as well, but I've noticed that once my queue pops and I enter whatever it is I'm doing, the latency to the server skyrockets. For example, I average ~55ms to the Malboro Server (199.91.189.24). I setup a continuous ping to the server and can see the latency is fine and I'm not experiencing any lag. As soon as my queue for trial roulette pops, the latency nearly doubles. At it's worst, I've seen it over 200ms

Fri Oct 3 19:21:18 EDT 2014 | 64 bytes from 199.91.189.24: icmp_seq=1 ttl=239 time=294 ms
Fri Oct 3 19:21:20 EDT 2014 | 64 bytes from 199.91.189.24: icmp_seq=1 ttl=239 time=93.2 ms
Fri Oct 3 19:21:21 EDT 2014 | 64 bytes from 199.91.189.24: icmp_seq=1 ttl=239 time=223 ms

It could be coincidence, but it seems almost like a trigger with how the latency and the queue popped at the same time. Yesterday the latency kicked in around 6:55pm EST on it's own - same time as listed above.

Unless AT&T is doing some QoS Traffic shaping, this seems a bit odd. Is there some other way I should be testing the latency? Are there different servers for the primal instances and such? I know many people have complained of this lag since the latest patch, but I've only just seen this behavior after the patch and after upgrading from their DSL 6Mbp/s line to the 18 Mbp/s U-Verse line. The same medium is used, so it seems likee more of a traffic shaping or routing issue.

docfiord@desktop:~$ tracepath -n 199.91.189.24
 1:  192.168.1.103                                         0.108ms pmtu 1500
 1:  192.168.1.254                                         1.026ms 
 1:  192.168.1.254                                         0.895ms 
 2:  External Next Hop                                   26.261ms 
 3:  76.201.208.74                                        25.545ms 
 4:  76.201.208.125                                       25.665ms 
 5:  12.83.82.145                                         27.000ms 
 6:  12.122.141.233                                       29.426ms asymm  7 
 7:  no reply
 8:  no reply
 9:  no reply
10:  no reply
11:  no reply
12:  4.69.134.70                                          53.181ms asymm 16 
13:  no reply
14:  4.69.141.5                                          140.592ms asymm 13 
15:  4.69.141.1                                           53.807ms asymm 14 
16:  4.59.178.74                                          53.472ms asymm 15 
17:  no reply
18:  192.34.76.2                                          59.605ms asymm 14 
19:  no reply
20:  no reply
21:  no reply
22:  no reply
23:  no reply
24:  no reply
25:  no reply
26:  no reply
27:  no reply
28:  no reply
29:  no reply
30:  no reply
31:  no reply
     Too many hops: pmtu 1500
     Resume: pmtu 1500

*EDIT*
To help with the troubleshooting, try to get a ping sample and a traceroute to your server from both when it is working fine and when there is lag. Here's how you can do it. If you are on a PS3 or PS4, doing this from your PC will still provide a semi-accurate reading as both devices are rpesumably using the same gateway.

Server IP addresses can be found from http://www.arrstatus.com/ by changing the Display Mode to IP Address. Copy and paste the reults from each test below.

PING
Windows XP : Start > Run > Type "cmd" (without the "") and press ENTER. Type "ping x.x.x.x" (substitue x.x.x.x for your server's actual IP address)

Windows 7: Start > Type "cmd" (without the "") and press ENTER. Type "ping x.x.x.x" (substitue x.x.x.x for your server's actual IP address)

TRACEROUTE
Windows XP : Start > Run > Type "cmd" (without the "") and press ENTER. Type "traceroute x.x.x.x" (substitue x.x.x.x for your server's actual IP address)

Windows 7: Start > Type "cmd" (without the "") and press ENTER. Type "traceroute x.x.x.x" (substitue x.x.x.x for your server's actual IP address)

[Hyperia]

Around 8:00 PM EST I start to get bad lag on U-Verse, its inbound lag though, outbound seems to be fine. I get 10 second pauses perodiclaly.

[Kassatsu]

I have the same probably usually start at 7:30pm-11pm est on UVerse as well. I live in Fl. Typically Only when I'm in instances. I get lots of a few seconds of lag then larger ones of like 30seconds.

[Zozon]

I've also been getting really bad lag on uverse (SoCalifornia) since this last patch. If it is AT&T messing with stuff like you suggest, it's just a preview of what's to come. Cross your fingers that it's something SE did with the patch.

[stellar_force]

I'm also getting huge lag spikes, and am on Uverse. Only been playing for about a month and when I first started playing it was fine. Now its every night especially in dungeons. I'm switching my ISP next week and will report back.

[Docfiord_Fowling]

Took a sample last night. The lag/latency began at 6:54pm EST and lasted until 8:18am EST this morning. I usually average 50-60ms of latency to the server, so I set a script to record anything over 80ms of latency. This still leaves 20ms of delay for overhead and anomolies. Looks like there is about a 13 hr window of latency. Will post tonight's results as well.

Fri Oct 3 18:54:53 EDT 2014 | 64 bytes from 199.91.189.24: icmp_seq=1 ttl=239 time=106 ms
Sat Oct 4 08:18:02 EDT 2014 | 64 bytes from 199.91.189.24: icmp_seq=1 ttl=239 time=96.7 ms

[Docfiord_Fowling]

Hrm. It appears there are also some random lengths of time throughout the day where this happens as well. Ran out to do some errands and came back to about 35 minutes of consistent latency. Perhaps SE is doing some maintenance/staging for the upcoming 2.4 release?

Sat Oct 4 14:05:55 EDT 2014 | 64 bytes from 199.91.189.24: icmp_seq=1 ttl=239 time=92.2 ms
Sat Oct 4 14:39:20 EDT 2014 | 64 bytes from 199.91.189.24: icmp_seq=1 ttl=239 time=109 ms

[Velocy]

Personally, first thing is: I question if arrstatus is REALLY showing the correct IP address. As an example, I play on Ragnarok. I see Ragnarok and Moogle have the same IP as for ARR Status.

I cannot ping my Server (timeout). However, using netstat while being logged in, I can see 2 IP adresses I am connected to. 199.91.189.106 and 144. None of them Show up on arrstatus.

Secondly: As far as I know or understand... instances are handled by completely different Servers within the datacenter than the world Server. I conculde that from the past. Remember the time where Chimera and Hydra were non-duty finder instances? That was a time where a LOT of People were Camping in front of the entries and tried to get parties. The Instances were lagging really really badly. That was the time where the World-Server handled the instance. Later, they shifted the instance to the instance Server(s?), and the common lag was gone. (And I remember I read something about that in some old patch notes).

That's also the best explenation to me, why you play with People from different Servers (but not from different datacenters). Also I might think there is more then one instance Server. From the time we did Titan Ex and HM with my static I remember times where almost everyone had Trouble getting out of plumes or landslides... exiting instance, reentering eventually fixed that.

So what you probably want to do is:
netstat -a | find "199.91.189" in a Dos Box, while normally connected, while in an instace to find the correct IP adresses.

If you are playing on PC, I would suggest to try out a vpn solution like pingzapper, battleping or wtfast. Some of them have 30-60 minutes Trials which should be enough time if your Problem can be reproduced by entering an instance or for a certain timeframe.

[Docfiord_Fowling]

Unfortunately, I only have the PS4 version, so I can't make use of the conenction state table. However, since you can, how about finding a few things for us?

Like you mentioned, check the state table and list a few entries you know to be an instance server. Please make sure it is showing the src and dst ports as well - I'm curious to see if I need to open any that are not right now.

[Velocy]

I'll gladly get back to it tomorrow.
Of course I can post the "Default ports" already:
UDP
55296-55551
TCP
80, 8080, 443, 55296-55551

Also one more thing I'd like to suggest: Check for traffic shaping by your ISP. The Max Planck Institute developed the Glasnost Test for that: http://broadband.mpi-sws.org/transpa...ttest-mlab.php
(especially run the test during your peek times). FFXIV uses TCP and UDP High Ports, which could be falsely recognized as P2P Traffic by either your ISP or a Sub-Provider and Drops to the bottom in priority.