GGG, you seriously need to do something about your international ISPs

"

neohongkong wrote:

post the WinMTR result to prove it

Not sure why you're so defensive. Here you go.

https://pastebin.com/NscsGMvx

Packet loss of 65% and 97% reported from several nodes starting at Telstra Global. This is followed by continuing failures at an ISP called "Root SA" starting at IP 94.242.252.229 followed by another ISP "Servers.com B.V." at 23.109.123.89. Whether these are actual ISPs or simply data centers, I don't know. I'm just doing a reverse IP lookup on them. But all of these IP addresses are based in Singapore.

Then it does not show the problem you claimed.

Instead it may be SG ISP own problem.


GGG did not own the trans-Pacific cable and there should be more than one but unsure your SG ISP use which one



Note that GGG only have control to the last IP in your record, and all the "telstra" are Telstra the Australia company

Like i said GGG have no control of how or which way you client routes to its server in your locale, that’s entirely under the remit of the ISP services. Only thing GGG could do is provide more servers for a locale/region and hope (there would still be no guarantee) your traffic is routed better and through less ‘poor performing’ nodes

"

neohongkong wrote:

Then it does not show the problem you claimed.

Instead it may be SG ISP own problem.


GGG did not own the trans-Pacific cable and there should be more than one but unsure your SG ISP use which one



Note that GGG only have control to the last IP in your record, and all the "telstra" are Telstra the Australia company

The letter T strikes again, Telstra in his region and Telia in mine lol

"

Timbo Zero wrote:

Only thing GGG could do is provide more servers for a locale/region and hope (there would still be no guarantee) your traffic is routed better and through less ‘poor performing’ nodes

Did you read my post? Because this claim is in contradiction to what I wrote. They do have other approaches. They can do something other than making new servers. If you disagree with anything in particular in my post please do point it out and explain why its not true or relevant.

"

Zrevnur wrote:

GGG has the power to significantly improve these issues.
2 obvious approaches:

1) Stick with good routes and ditch bad routes. One of the symptoms in EU (for me at least) is that some connections are much better than other connections. There is no inherent need to reconnect all the time like GGG does. And if they want to stick with their reconnect-all-the-time scheme they could still do better by ditching bad routes per building another connection. Obviously this wont fix everything. But almost all the bad connections in this league which I had (in EU) were of this type that "manual reconnection" (leaving area and reentering) had a good chance to fix it.

2) Make/use proxy (not sure this is the right word) servers against known bad routes.

(There is probably more they can do but I have limited technical insight.)

1) the routing between you (your game client) and them (in the case of console and EU this is either London or Frankfurt, there are additional ones for PC such as I believe Paris etc..) is entirely out of the control of GGG, they can do nothing other than 2).

2) your game does not connect to ‘GGG’ but it connects to (usually, but not exclusively as other options exist) an AWS or Azure server ‘blade’ in the locale (I.e. as per above London or Frankfurt). They could indeed as you have said add more of these server ‘blades’ in locales that would reduce traffic through ‘undesirable’ ISP services that are ‘known’ to cause issues especially at peak times. So for console a Paris , Poland etc for example MIGHT help but does NOT guarantee the issue will be solved or improved because they still have no control over how the ISP services route you.

As an additional note, you mentioned the reconnect issue. POE seems, can’t be sure without deep diving their code and I don’t have time atm to do so, to ‘double hand’ some tarfffic as part of their ‘anti cheating’ policy. For example your position as the player, whereas other assets are essentially handled by the game client and then ‘verified’ by the server. The game isn’t ‘reconnecting’ but basically requires 100% time connected to the server, this is inefficient but it’s how they’ve chosen to do it :(

"

Timbo Zero wrote:

1) the routing between you (your game client) and them (in the case of console and EU this is either London or Frankfurt, there are additional ones for PC such as I believe Paris etc..) is entirely out of the control of GGG, they can do nothing other than 2).

Maybe I was unclear before. Trying again:
The situation which I have experienced (>90% of my issues this league) and which has been described in various reddit posts which I have read - most of them for EU but not all:

"

1) I enter map. Map/Instance is laggy.
2) I leave map and reenter it. Now I have a 50% chance of map being good and a 50% chance of map/instance being laggy.
3) If its laggy goto 2). If its not laggy play map.
Note that in almost all cases the map is either consistently laggy or consistently good. This (it being laggy or good) can be "rerolled" by leaving and entering. The chance for it be laggy was different. Usually this league it was 0%. But there were bad times of varying degrees. At the worst times this chance was >50%.

So from that I concluded that reconnecting to that instance can fix the problem. My (not really) understanding here is that not all connections between the same client and the same server/instance have the same routing. Meaning if multiple connections are built in sequence some may have different routings. And some routes are better than others. Consequently GGG can affect the routing by closing connections with bad routing and trying again. Far as I understand this is what happens when I leave map and reenter it.
Now my point here is that instead of me "manually" doing this GGG can automatically do it.

"

Timbo Zero wrote:

They could indeed as you have said add more of these server ‘blades’ in locales that would reduce traffic through ‘undesirable’ ISP services that are ‘known’ to cause issues especially at peak times.

I dont understand the technical details here. So not sure if this is what I meant with "proxy". With "proxy" I basically meant to do what a VPN service does. (It has been recommended (due to positive experiences) to use VPN services.)

"

Timbo Zero wrote:

The game isn’t ‘reconnecting’ but basically requires 100% time connected to the server, this is inefficient but it’s how they’ve chosen to do it :(

See above - not closing good connections is IMO a good thing.

"

Zrevnur wrote:

"

Timbo Zero wrote:

1) the routing between you (your game client) and them (in the case of console and EU this is either London or Frankfurt, there are additional ones for PC such as I believe Paris etc..) is entirely out of the control of GGG, they can do nothing other than 2).

Maybe I was unclear before. Trying again:
The situation which I have experienced (>90% of my issues this league) and which has been described in various reddit posts which I have read - most of them for EU but not all:

"

1) I enter map. Map/Instance is laggy.
2) I leave map and reenter it. Now I have a 50% chance of map being good and a 50% chance of map/instance being laggy.
3) If its laggy goto 2). If its not laggy play map.
Note that in almost all cases the map is either consistently laggy or consistently good. This (it being laggy or good) can be "rerolled" by leaving and entering. The chance for it be laggy was different. Usually this league it was 0%. But there were bad times of varying degrees. At the worst times this chance was >50%.

So from that I concluded that reconnecting to that instance can fix the problem. My (not really) understanding here is that not all connections between the same client and the same server/instance have the same routing. Meaning if multiple connections are built in sequence some may have different routings. And some routes are better than others. Consequently GGG can affect the routing by closing connections with bad routing and trying again. Far as I understand this is what happens when I leave map and reenter it.
Now my point here is that instead of me "manually" doing this GGG can automatically do it.

"

Timbo Zero wrote:

They could indeed as you have said add more of these server ‘blades’ in locales that would reduce traffic through ‘undesirable’ ISP services that are ‘known’ to cause issues especially at peak times.

I dont understand the technical details here. So not sure if this is what I meant with "proxy". With "proxy" I basically meant to do what a VPN service does. (It has been recommended (due to positive experiences) to use VPN services.)

"

Timbo Zero wrote:

The game isn’t ‘reconnecting’ but basically requires 100% time connected to the server, this is inefficient but it’s how they’ve chosen to do it :(

See above - not closing good connections is IMO a good thing.

OK, so I understand a little more about your ‘misconceptions’ if you will forgive the term.
When you enter a new ‘instance’ and re-entering a map can cause this (it doesn’t need to be a new map) you create a new connection to the local server blade and this new connnection might route differently to the previous one and avoid (you are on PC so sorry I have to use XBox examples here) the renowned Telia ISP that causes EU issues and why most of console EU players use London server over Frankfurt server (to avoid or improve chances of avoiding ‘bad’ ISPs.)
The routing is not done by GGG or even slightly within their control, it is a combination of factors including local loads on ISP nodes and servers, deals of service between various ISPs and loads upon ‘backbone’ services at any given second.

You mentioned a VPN, this would have next to no impact on the issue. Your game would still need to route to the VPN (although a good route would be maintained, as would a bad route) and then from the VPN server to the local GGG server which would be as variable as before but hypothetically over a shorter distance.
Again this is a lottery call and the route is still out of GGGs hands.

The not closing good connections is sensible but it would also mean not closing bad ones either as GGG doesn’t know you have a good or bad one. Making a new connection each time you start a new activity is a craps shoot either way sadly.

The only real cure is for the problematic ISPs (not GGG) to improve their infrastructure and with any luck the EU directive to make us all have fibre by the end of 2025 (iirc) should help a lot with this potentially

Footnote , apologies as I am not explaining this particularly well. I’m 58 and haven’t worked in the industry for 20 years, it’s early here and haven’t finished my morning coffee, and I’m even having to recheck posts for typos 3 times lol

"

Timbo Zero wrote:

The routing is not done by GGG or even slightly within their control, it is a combination of factors including local loads on ISP nodes and servers, deals of service between various ISPs and loads upon ‘backbone’ services at any given second.

The important part is that its potentially different.

"

Timbo Zero wrote:

You mentioned a VPN, this would have next to no impact on the issue. Your game would still need to route to the VPN (although a good route would be maintained, as would a bad route) and then from the VPN server to the local GGG server which would be as variable as before but hypothetically over a shorter distance.

My VPN part was based on (many, anecdotal) claims that using a VPN helps or completely fixes the issues. Why that is is from my POV open to speculation. Anyway this (VPN working or not) has no bearing on the other part (reconnect to kill bad routes).

"

Timbo Zero wrote:

The not closing good connections is sensible but it would also mean not closing bad ones either as GGG doesn’t know you have a good or bad one.

If I the player can tell if a connection is good or bad then an algorithm by GGG can tell - and probably much better because it can actually properly see the issues (like packet losses). So yes GGG can distinct between good and bad connections.
And as I wrote in another thread - even GGG providing a "/reconnect" command would be better than nothing. Especially in cases where reentering isnt possible.
See older thread (by me) for another take on this: https://www.pathofexile.com/forum/view-thread/3555268

distributed system architect here

there are two potential issues at play here, both costly to fix (not hard, costly)

1) local instance instability. poe uses instances (thin poe 'clients' that run chromeless, by dozens per standard DC node). these instances are provisioned based on statistical data of prior load and usage (as instances are short lived i doubt they do live up/downscaling, so it is pre-planned and instances are pre-spun). this is done to maximize the number of instances per node. there are cases - like VERY laggy builds being popular in given league - where this statistical approach results in instances being disturbed by so called 'noisy neighbour' instances. simplified example: there are 10 instances per node. 'normal' players are using around 9% of node cpu cycles with 10% margin used for peaks (instance load is VERY non-linear). everything is fine. but if one of these players uses some kind of ward-loop idiocy with 500 casts per second and faces expedition with mobs invulnerable to the damage type it will use 30 or more % of node power STEALING it from other players. this is very noticeable for players affected. real scenario most likely operates at 50-100 instances per node
in theory it can be provisioned for - making the buffer larger or allocation logic smarter - but it costs. hosting costs are SIGNIFICANT and im sure GGG is cutting corners wherever possible. there is a reason league starts are what they are (ignoring bugs for a moment)

re-entering map MIGHT fix it, if the original instance is closed and you get loaded into newly spun on a different node. but if the problem is not severe, orchestrator might keep assigning you to the same instance/node. it is also possible that youll get 'couldnt join any instances' if the instance is already dropped/crashed and its state wasnt synced before so the backup couldnt be created

2) routing. various issues are in play here, im familiar with just some of them - but there is no secret in traffic being cut unevenly and some players having better access to busy routes than others (big players in general either use their own hardlines or pay to even bigger players for exclusivity). POE is a minor player, so small that i bet it gets the worst possible prio everywhere possible. i doubt GGG has any leverage here. there are known issues at connection points between various big networks, there are maintenance breaks of accidents (one forest fire somewhere remote could result in entire DC being downgraded due to to damaged power supply backup - these issues ripple and multiply like crazy

it is possible to request certain routings, use internal ISP landlines if possible etc - but i bet each choice has a hefty price tag attached that only the big peckers can accept.

in short - small players cannot really do much and POE despite what many here think is marginal player with pretty much no negotiation leverage. i think that the most realistic path - that GGG wont take - is to hard-limit certain idiotic mechanic in their game (like self-harm loops, cast when stunned etc) - these overwhelm their engine and make any form of load prediction guesswork. and that can only be fixed by some serious OVERprovisioning. given that GGG doesnt want to pay for it we get what we get each league 'the diarrhoea' builds are popular