Significant lag

[Chrsitin]

Verizon FIOS. Significant lag for the past 24 hours.

[VertexGeneral]

Significant lag on almost every provider for over a month.

[SooChe]

I would like to verify this. I have been compromising on joy for the past month. Lag and worse lag day by day. Why is SE so rich yet have shirty servers. Joy turned to frustration turned to despair, turned to abandonment....

[Oopsy]

Check the stats http://forum.square-enix.com/ffxiv/t...nd-disconnects

This is not an SE issue, packet loss is happening on Level3 communications servers as well as high latency.
Most is happening once it hits Montreal occasional hick ups on my isp's side, not often but they happen.

******************************WARNING WALL OF TEXT******************************

How to find the issue?
UOTrace is a easy tool. This tool is made by EA for their ultima online servers. Silly ultima online... people often over look it... One of the first major mmo's... Well gaming history aside this tool makes figuring out where the issues start easy. When you find the official download on EA's website it will have instructions on how to use it.

For advanced users:
Tracert - this is a command built into windows computer it will find the path you take to a server.
Pathping - this is a command built into windows computer it will find packet loss over a route to a server

Once you get some results how to read them and what they mean is the hard part.

Here's some terms and an explanation:

Packet loss:
Loss of a package normally read as a percentage.
When sending information over a network it is broken down in to "Packets" that range in size between 256 bits and 1500 bits. Your normal picture on a cell phone is about 3mb or 24576 bits so to send that over a network you would want to break it down into 17~96 packets. Imagine how bad a picture would look if 10% of the picture was missing? I would think ffxiv sends packets of rather large size most games do.

Latency
How long it takes for information to be received from when it is sent. Represented as Ping measured in milliseconds 1/1000th of a second. A round trip.
Imagine you are watching a foot ball game on a TV at a stadium. Your favorite player just scored it was amazing! If you started a timer when they scored and stopped it when the tv showed that score that would be the latency. How late something is.
If you only know about the aoe 1 second after it appears you will never be able to properly dodge it.

What control does SE have over these?
Well... Some and yet very little. The last few hops at most 3 or 4 hops at most.

What do you have control over?
A lot actually... If anything is happening in the first hop or in a hop with an ip address starting with 192.168... or 10.0.... These are likely your home network. After that when things start having a domain (website name .com .net ect) they are going start in your internet providers network look for the name of the company, it maybe a bit abbreviated so keep an eye on out for it. The internet provider is a network you have sway over. Call you ISP with stats you have found and prove that they are causing packet loss or high latency in their network they need to "check the lines" or send out a technician.

Why?

Light travels at a specific speed and can not travel faster.. the same goes for electricity though its only 1/100th the speed of light. This means depending on your distance to the server you can not get less then a certain latency. SE can make sure they are on a fiber optic connection reducing the latency by using light rather then electricity for as far as possible but if you are not on a fiber connection there will be some where it just cant get the extra bump... no worries though it wont be long till you hit the "Back Bone of the internet" then it will be fiber till you hit your goal. If you do a trace route and then go to google maps travel from home to Montreal CA and add any city name you see in your route to the map you will get a ruff estimate of the distance your information has to travel. Now take that and divide by 185 to ruffly estimate the time in milliseconds it takes light to travel that far. This will be the absolute minimal time it will take for your information to reach the server double this and you got your minimal ping. This is assuming no computers have to look at the data to forward it. If you add 1~2 ms to your ping for each hop on a trace route you would be having a perfect connection.

Below is a sample from my own issues connecting to the server currently.

Host Name                                                             IP Address        Hop    Ping Time   Ping Avg   % Loss    Pkts r/s             Ping best/worst
router.dlink.com                                                      192.168.0.1       1      2ms         0ms        0%        25 / 25              0ms / 2ms
removed.frontiernet.net                                                 removed         2      5ms         4ms        0%        25 / 25              2ms / 15ms
ae11---0.car01.bvtn.or.frontiernet.net                                74.40.70.41       3      5ms         5ms        0%        25 / 25              2ms / 14ms
ae2---0.cor01.bvtn.or.frontiernet.net                                 74.40.1.217       4      3ms         9ms        0%        25 / 25              2ms / 41ms
ae4---0.cor01.sttl.wa.frontiernet.net                                 74.40.1.221       5      8ms         21ms       0%        25 / 25              8ms / 107ms
ae0---0.cbr01.sttl.wa.frontiernet.net                                 74.40.5.122       6      9ms         13ms       0%        24 / 24              7ms / 35ms
ae57.edge1.Seattle3.Level3.net                                        4.59.232.5        7      29ms        34ms       4%        23 / 24              23ms / 67ms
ae-10-10.car2.Montreal2.Level3.net                                    4.69.153.86       8      86ms        94ms       0%        24 / 24              83ms / 188ms
ae-10-10.car2.Montreal2.Level3.net                                    4.69.153.86       9      86ms        97ms       0%        22 / 22              83ms / 180ms
ORMUCO-COMM.car2.Montreal2.Level3.net                                 4.59.178.74       10     142ms       109ms      13%       19 / 22              83ms / 166ms
* Unknown Host *                                                      192.34.76.10      11     154ms       95ms       4%        21 / 22              80ms / 154ms
* Unknown Host *                                                      199.91.189.242    12     90ms        98ms       0%        22 / 22              85ms / 174ms
* Unknown Host *                                                      199.91.189.37     13     85ms        86ms       0%        20 / 20              83ms / 114ms

So based off the information above I have 13 hops and as you can see my the host names my connection starts in Bvtn or... Short for Beaverton Oregon and travels north to sttl wa.... Short for Seattle Washington. Then hops on the internet back bone off to Montreal. At this point I am no longer in my internet providers network as you can tell by the switch from Frontiernet to Level3.

With a quick look on google maps I get just over 3000 miles.
3000/185 is 16.2 milliseconds to tell the server what I want. I then have to wait on a reply so another 16.2 for the way back. This is 32.4 milliseconds as a 100% perfect connection not even pausing to take to the servers on the way. So lets add in the server latency that would be normal if not too good to ask for. 13 hops x 2 = 26 ms one way and another 26 add that to the 32.4 and we got 84.4 milliseconds. Pretty darn close to my best ping on hop 13 (SE's server). The math is sound.

But there was 4% packet loss on hop 7? and 13% at hop 10?
Now one thing to remember... Each hop has to go through each hop before it. When looking at packet loss you must look at the earliest hop. It is possible the packet loss on later hops are the earlier hops loosing the packet sent. If you were to pass a ball down a line of 10 people. Each person passing it to the next and are trying to got person 1 they then put it in a basket. then person 2... person 3.... but then on person 4 they drop the ball... well you keep going and pass to person 5... and 6.... but when passing to person 7 the 4th person's slippery fingers drop the ball again... well that still counts as person 7 losing the ball. Packet loss does not know if a person they were not testing failed to pass the ball along it just knows the person they were testing did not replay.

uotrace and path ping will show "Pkts r/s" or equivalents. these are the number of packets you sent (s) as a test and how many were revived (r). The packet loss is not very accurate till you get to about 200 packets sent.

the max ping on hop 8 is 188ms that's pretty high. A normal person will notice latency above 150ms while playing video games 200ms while watching a video and 300ms while loading a website. For a smooth gaming experience I recommend a ping under 100.

Again there is a trick to this one as well... Some server prioritize certain traffic very low... These servers often put ping tests on a low level..
Luckly this uotrace has a "Ping Avg" and "Ping best/worst" that way you know if things are normally high or if a few spikes just happened to get mixed in and what the best connection you got were. If the best is close to the avg the connection is pretty stable and the server is not likely putting pin at a low priority.

Then again... It is my job to prove network issue are not happening server side for a console's network so you should take my information with a grain of salt. My thought process is to find the issue that is happening out side the end point. Things the customer can change and control. The network admins are already working full time to identify and resolve any issues they can if you can find something your side to fix we can meet in the middle with a much better connection all around.