Terrible bufferbloat, packet loss, jitter, etc. Text support service DOES NOT WORK

[#1]

Ok I've got WinMTR. What am I doing with it? 

Atm I am getting spikes even with the wifi off, though not as bad as when it was on. This is to a UK server.

|------------------------------------------------------------------------------------------|
| WinMTR statistics |
| Host - % | Sent | Recv | Best | Avrg | Wrst | Last |
|------------------------------------------------|------|------|------|------|------|------|
| 192.168.1.254 - 0 | 84 | 84 | 0 | 1 | 8 | 0 |
| 172.16.10.220 - 0 | 84 | 84 | 6 | 7 | 10 | 7 |
| No response from host - 100 | 16 | 0 | 0 | 0 | 0 | 0 |
| 140.hiper04.sheff.dial.plus.net.uk - 0 | 84 | 84 | 8 | 9 | 13 | 9 |
| peer7-et-4-1-5.telehouse.ukcore.bt.net - 2 | 80 | 79 | 8 | 13 | 73 | 9 |
| be107.lon-thw-pb1-nc5.uk.eu - 0 | 84 | 84 | 8 | 9 | 14 | 8 |
| No response from host - 100 | 16 | 0 | 0 | 0 | 0 | 0 |
| No response from host - 100 | 16 | 0 | 0 | 0 | 0 | 0 |
| lon1-eri1-vac1-a75-1-firewall.uk.eu - 0 | 84 | 84 | 8 | 9 | 12 | 9 |
| lon1-eri1-vac1-a75-3.uk.eu - 0 | 84 | 84 | 9 | 9 | 14 | 10 |
| No response from host - 100 | 16 | 0 | 0 | 0 | 0 | 0 |
| No response from host - 100 | 16 | 0 | 0 | 0 | 0 | 0 |
| No response from host - 100 | 16 | 0 | 0 | 0 | 0 | 0 |
| ns3207745.ip-198-244-177.eu - 0 | 84 | 84 | 8 | 9 | 12 | 9 |
|________________________________________________|______|______|______|______|______|______|
WinMTR v0.92 GPL V2 by Appnor MSP - Fully Managed Hosting & Cloud Provider

|------------------------------------------------------------------------------------------|
| WinMTR statistics |
| Host - % | Sent | Recv | Best | Avrg | Wrst | Last |
|------------------------------------------------|------|------|------|------|------|------|
| 192.168.1.254 - 0 | 123 | 123 | 0 | 0 | 8 | 1 |
| 172.16.10.220 - 0 | 123 | 123 | 6 | 7 | 10 | 8 |
| No response from host - 100 | 24 | 0 | 0 | 0 | 0 | 0 |
| 140.hiper04.sheff.dial.plus.net.uk - 0 | 123 | 123 | 8 | 9 | 13 | 9 |
| peer7-et-4-1-5.telehouse.ukcore.bt.net - 1 | 119 | 118 | 8 | 12 | 73 | 22 |
| be107.lon-thw-pb1-nc5.uk.eu - 0 | 123 | 123 | 8 | 9 | 16 | 10 |
| No response from host - 100 | 24 | 0 | 0 | 0 | 0 | 0 |
| No response from host - 100 | 24 | 0 | 0 | 0 | 0 | 0 |
| lon1-eri1-vac1-a75-1-firewall.uk.eu - 0 | 123 | 123 | 8 | 9 | 12 | 10 |
| lon1-eri1-vac1-a75-3.uk.eu - 0 | 123 | 123 | 9 | 9 | 14 | 10 |
| No response from host - 100 | 24 | 0 | 0 | 0 | 0 | 0 |
| No response from host - 100 | 24 | 0 | 0 | 0 | 0 | 0 |
| No response from host - 100 | 24 | 0 | 0 | 0 | 0 | 0 |
| ns3207745.ip-198-244-177.eu - 0 | 123 | 123 | 8 | 9 | 12 | 9 |
|________________________________________________|______|______|______|______|______|______|
WinMTR v0.92 GPL V2 by Appnor MSP - Fully Managed Hosting & Cloud Provider

Edit: I see it doesn't paste too cleanly. Here's a screenshot, I should've thought of that sooner. Sorry.

I'll then try it again later if this doesn't show anything obvious.

[#2]

Looking at the Broadband Quality Monitor, something that is usually "chatty" on the network was disconnected between about 7:15pm and 8pm.

My guess is that's the device (or devices) that cause increases in the maximum latency (yellow on the graph).  Generally very little happens to the average (blue).

My guess is that the Plusnet Hub responds to a WAN (Internet) ping as it's lowest priority so any traffic flowing delays the ping response.

While a longer (peak time) period of everything disconnected would be useful, I tentatively conclude that the connection is probably behaving "normally".

My guess (and it's just that) is that the gaming requires quite a lot of traffic and is very delay sensitive. Without knowing more about the game server locations (which may have moved or the routing changed) it's hard to say why the problem started a month ago.

What I don't see is evidence of a Plusnet "fault" that can be fixed.

The streaming issues may well be WiFi related.

[#3]

Dear Mark,

Thank you for your reply.

That was when I turned off the whole wifi, yeah. However, for large parts of the day (with the wifi on), the spikes were bad even with nothing on the wifi (though it was on). 

I will have the wifi disconnected today for most of the day, so hopefully that'll help. There have been quite a few spikes even when the wifi was turned off, but it seems to be settling down a bit now (I turned it off maybe 1 hour ago). The problem with the 'normal' hypothesis is that my qualitative experience with the latency has declined a lot in the last month or so. Before, my connection was always pretty much fine, even when the wifi was on and there was a device connected to it. It is only recently that it has become near-unuseable even when there aren't devices actively using the wifi.

And looking at the numbers the actual amount of data downloaded/uploaded (I have the numbers for dl, I can estimate for ul but I will ahve proper numbers tonight) is miniscule and not even 1/100th of the way to saturating the connection.

So there must be a more recent change in my connection to have caused the issue. Yet there isn't a change on my end that could've caused this, nor on the end of the game (I am getting these issues to every address, not just in the game, e.g., if I ping bbc.co.uk in cmd I get the same issue, and so on and so forth).

So there must be an issue that is exacerbating this and causing it to 'saturate' even when the wifi is barely being used, right? People have said upgrade and, yeah, I know the deal I have sucks (I don't control it, I am renting atm--my landlord is paying £37(!!!!!!!!) a month for 40/10(!!!!!!)), but the point is that it worked fine until very recently such that the bad router/speeds are not the primary causal mechanism here because they have remained constant throughout the good and bad period. 

Unfortunately I don't have a pre-issue BQM, so maybe it would look the same regardless--no idea, but certainly the qualitative experience (and the quantitative jitter in use) has gotten way way way worse.

[#4]

I get what you are saying, and something seems to have changed.

If you look at the "explainer" for the Broadband Quality Monitor

https://www.thinkbroadband.com/guides/broadband-quality-monitor

you can see what "bad" connections look like.

Your Plusnet connection doesn't look like any of the typical faults.

Don't get over concerned with the occasional maximum latency spikes.

The MTR (much easier to interpret as a picture) looks "normal" and very similar to the trace route I just ran here. Is the endpoint of relevance to your gaming? (it looks like an Apache server holding page)

The only packet loss was on BT's peer at Telehouse (hop 5). That also shows a "worst" of 73ms which suggests it was fairly busy.

Someone else (an EE customer but it's the same network) was chasing a similar issue recently.

This is worth a read and mentions the next hop in your trace route.

https://community.ee.co.uk/t5/Broadband-Landline/High-Jitter-and-Packet-Loss-on-EE-Backhaul-Nodes-Ho...

Looking for the thing that might have changed, I'm guessing the routing to the game you are using. There are many reasons why routing changes.

Perhaps worth a try would be some sort of VPN on your PC. Then the traffic would exit somewhere else (at the end of the VPN tunnel) but this could well cause more problems than it solves!

[#5]

So I performed a mtr on a linux system connected directly to my router, I have Full Fibre 900

What this details is the organisation owner of the IP Address

AS6871 = plusnet

AS2856 = BT

AS16276 = OVH cloud

For me the average ping is 6.6.

It would be interesting to have mtr running when you start and play the game.

Edit: A wifi mtr on my win 11 laptop

[#6]

Dear Dan,

Thanks, what server should I do it to? Just any old one? The official servers don't give you their IP as far as I know.

The one I posted above is to a 'community server' in the game but it's the same no matter what address I use, e.g., the bbc.co.uk address produces the same pattern of a decent average (with the wifi off, as aforementioned--I'll turn it on at like 8pm tonight) but with fairly regular spikes, some going very high. But these are always at hops that are quite a way away from me in the Plusnet/BT controlled nodes. Sorry, I don't really know how it works or what a 'hop' is in this sense, but I assume it means a router type thing in a building they own?

I'm curious to see that my BQM is showing pretty much the same pattern of spiking as it did when the wifi was on (it's off atm), so myabe that's not even the main solution? It's weird because it did produce a fairly stable environment last night, yet now with the wifi off (at a time of lower nation-wide traffic) it's worse? My connection to the game seemed better this morning but not perfect (good enough though), I'll try it again in a bit to see if it's unstable again.

[#7]

Dear Mark,

Thank you for your informative and helpful reply, I appreciate your patience with me.

The one I posted above is to a 'community server' in the game but it's the same no matter what address I use, e.g., the bbc.co.uk address produces the same pattern of a decent average (with the wifi off, as aforementioned--I'll turn it on at like 8pm tonight) but with fairly regular spikes, some going very high. But these are always at hops that are quite a way away from me in the Plusnet/BT controlled nodes. Sorry, I don't really know how it works or what a 'hop' is in this sense, but I assume it means a router type thing in a building they own?

I'm curious to see that my BQM is showing pretty much the same pattern of spiking as it did when the wifi was on (it's off atm), so myabe that's not even the main solution? It's weird because it did produce a fairly stable environment last night, yet now with the wifi off (at a time of lower nation-wide traffic) it's worse? My connection to the game seemed better this morning but not perfect (good enough though), I'll try it again in a bit to see if it's unstable again. Does the BQM even matter here, then?

On the BQM interpretation page, for irregular yellow spikes, it says:

"BQM graph showing the effect of hosting (upload) and browsing (download) on latency

The graph above was from a connection, that was hosting a small webserver, and the yellow spikes represent when someone is downloading a small file or webpage from the server. The wider block of yellow spikes at 11:45am represent when the actual owner of the connection, probably started using the connection too.

The connection in this graph is good, as the amount of jitter in the latency represented by the blue area is still very small across the whole 24 hour period.

If you are on a slow connection the action of downloading or uploading lots of data can have even larger effects, e.g. a connection that is saturated by streaming two HD films at the same time will most likely show a large yellow block for the duration of the streaming."

But the yellow spikes do not represent me downloading anything of note, at least, not consciously. I'm sure there is data being downloaded and uploaded, but as aforementioned the rate of download/upload (I can get the precise numbers if you give me a couple hours to track it, but I did it informally last night) are so low that they're not even 1/100th of the actual capacity of the connection. Why would it be that the connection gets saturated at even, say, 0.02% megabits/second dl and maybe 0.001 (guesswork but it's something like this) megabits/second upload? 

I'd get this if I were downloading something big while I was playing, but I'm not, of course.

----------------------------

Your hypothesis about routing makes sense, but I don't think it can be just to the game because the pattern is replicable across literally all addresses, even ones unrelated to the game. It seems every time when you do tracert or WinMTR or whatever it's called the spikes primarily and most commonly come from these hops (nodes/hops? IDK) that are at a distance from me and named something to do with Plusnet or BT. 

So the routing that my router uses as a whole, to anything? How do I change that?

As for VPNs, I'm not sure if you can get a free one that isn't extremely slow/laggy in itself. Not sure if there are any recommendations, I'm not familiar with the whole thing as atm my only VPN is just a browser extension on Chrome to get around Online Safety Act stuff e.g., Imgur haha.

But I'm happy to try if there's a good way to set it up without having to pay the VPN for decent speeds.

[#8]

I'm curious to see that my BQM is showing pretty much the same pattern of spiking as it did when the wifi was on (it's off atm), so myabe that's not even the main solution? It's weird because it did produce a fairly stable environment last night, yet now with the wifi off (at a time of lower nation-wide traffic) it's worse? My connection to the game seemed better this morning but not perfect (good enough though), I'll try it again in a bit to see if it's unstable again. Does the BQM even matter here, then?

On the BQM interpretation page, for irregular yellow spikes, it says:

"BQM graph showing the effect of hosting (upload) and browsing (download) on latency

The graph above was from a connection, that was hosting a small webserver, and the yellow spikes represent when someone is downloading a small file or webpage from the server. The wider block of yellow spikes at 11:45am represent when the actual owner of the connection, probably started using the connection too.

The connection in this graph is good, as the amount of jitter in the latency represented by the blue area is still very small across the whole 24 hour period.

If you are on a slow connection the action of downloading or uploading lots of data can have even larger effects, e.g. a connection that is saturated by streaming two HD films at the same time will most likely show a large yellow block for the duration of the streaming."

But the yellow spikes do not represent me downloading anything of note, at least, not consciously. I'm sure there is data being downloaded and uploaded, but as aforementioned the rate of download/upload (I can get the precise numbers if you give me a couple hours to track it, but I did it informally last night) are so low that they're not even 1/100th of the actual capacity of the connection. Why would it be that the connection gets saturated at even, say, 0.02% megabits/second dl and maybe 0.001 (guesswork but it's something like this) megabits/second upload? 

I'd get this if I were downloading something big while I was playing, but I'm not, of course.

----------------------------

Your hypothesis about routing makes sense, but I don't think it can be just to the game because the pattern is replicable across literally all addresses, even ones unrelated to the game. It seems every time when you do tracert or WinMTR or whatever it's called the spikes primarily and most commonly come from these hops (nodes/hops? IDK) that are at a distance from me and named something to do with Plusnet or BT. 

So the routing that my router uses as a whole, to anything? How do I change that?

As for VPNs, I'm not sure if you can get a free one that isn't extremely slow/laggy in itself. Not sure if there are any recommendations, I'm not familiar with the whole thing as atm my only VPN is just a browser extension on Chrome to get around Online Safety Act stuff e.g., Imgur haha.

But I'm happy to try if there's a good way to set it up without having to pay the VPN for decent speeds.

[#9]

More really good questions there.

The Think Broadband "explainer" was written from the perspective of an end Internet connection user (like yourself).

The bit that says "downloading a small file or webpage from the server" is assuming that the server is at the user's location so "download" (by someone else) equates to an upload from that location. Most UK broadband connections (yours included) are a faster "download" (to the user) than "upload" so if you are sending data (like a phone syncing photos) it uses the upload capacity and makes the response to the monitor's ping wait (hence the yellow spike of maximum latency).

Very occasionaly the router (or one of the other "hops" in the path between Think Broadband's test setup in Telehouse and the target). drops a response to the ping (or the incoming ping) and a lost packet is recorded.

A "hop" is an active routing device. These are large, expensive and very capable devices (often made by Cisco, Juniper or similar) that determine where Internet traffic should go.

Within BT's network the best path is often obvious and so you will see BT's devices closest to you. For big traffic generators (Netflix and similar) there may well be a store of their content on BT's network so it can be delivered in a reliable and controlled way and lots of effort is put into making such things work.

Your game server, on the otherhand, lives somewher outside BT's network. Of the hops that announce themselves "Faraday" says it is in the BT UK core, your traffic then goes to somewhere that BT's network meets others (it looks like Telehouse West at a guess) and then to the device which has "firewall" in the name.

This seems to be in Valve's system (search on the name for a bit more info). As I understand it (I'm afraid the last time I played a computer game it came on a large floppy disk) Valve and Steam are popular game providers. 

I suspect the remainder of the route is in a data centre in Erith.

There is one other thing to understand.

When you visit a text website, it generally uses "Transmission Control Protocol" (TCP). This checks for errors and if a packet of data gets lost or damaged it asks for a replacement. This is good but slow.

Applications that need lots of data (streaming, games etc.) often use the User Datagram Protocol (UDP). This is faster but it's "fire and forget" and it's down to the application (streaming client, game or whatever) to manage data (packet) loss.

Lots of errors get concealed (use the last good data for that bit of the picture as an example) but eventually the application isn't able to hide the problems from the user.

This is a very long way of saying that the Internet isn't designed to be perfect (it can't be) and it's down to the person designing the application to cope with that problem.

I can't fix your problem, and I'm not sure anyone can but I hope this helps to understand why things are as they are. 

[#10]

Here's one with the game on (to the very server), but with the wifi off. It felt largely ok, a few smaller spikes but nothing unbearable. I checked the downloaded data and uploaded data for this period and it was tiny on both sides.

So I'd suspect this one is at least a bit functional? It's to a UK server.

Then I'll try one tonight with the wifi on and my fiancee's phone attached to it, hopefully to the same server (for the sake of analysis), to see if there's a clear and obvious difference.