Are there any plans for Asia server?

Chromafunk has done a phenomenal job of FIXING this game's performance for both of us. It's true that I didn't have these performance issues in Diablo 3 or EverQuest. The quality of GGG's coding is inferior, I suppose. At any rate, as long as we can get past the Great Firewall of China, the game flies for us now.

"

Hey there,

I just wanted to clear something up when using WinMTR program. When looking for packet loss, you need to check the "sent" column in the first line vs the "received" line in the last line.

In your first trace, 255 packets were sent, and 251 were received by the server. This means that out of the whole trace, you suffered somewhere around 1% packet loss.

Hops where you see one big lump of packet loss, and then normal behaviour after are likely to be victims of "data shaping", which is essentially the hops giving WinMTR packets lowest priority. Your actual game data wouldn't be getting lost, just the trace route packets (which use a different kind of packet, and thus can be ignored by the hosts).

Actual packet loss should show up through the whole trace. IE, 80% packet loss would be something like 100 packets sent in the first line, and only 20 packets received in the final line. You would see consistent packet loss throughout the entire trace.

Hi Brian, thank you for showing up. I know how to look at the packet loss, what I wanted to say is that at that exit point in the Beijing network : "broad.bj.dynamic" is where the bottleneck is going on, actually that picture does not reflect the actual packet loss. normally its much more insane, but 250 takes is not enough to reflect it. I was stressed out doing the tests and working on the server and forgot to leave it tracing more time.

What is clear is that there is a point on your network side that should be revised, though. The hop "po2.fcr01.sr03.sng01.networklayer.com" causes packet loss. A switch interface? A router? At least would be cool to know why, as its on your side !

I will post more data now to you so you can see some interesting stuff :

-Pinging my public IP ( not locally )VPN server in Singapore from my computer going through the VPN tunnel ( I have switched the implementation to a Layer 2 bridge instead of using S-NAT, the performance and stability is substantial )




-Pinging sg.login.pathofexile.com also from my computer through the VPN tunnel



Evidence :

Given the fact that we have technically circumvented part of the deep packet inspection, dropping and blocking when we exit China ( There is no way of doing it entirely as there are also tons of physical network layer issues for sure )

My server is in Singapore, and your server too.

My server has a max burst of 240 ms. Your server has almost constant bursts of 800+ ms being on the same place geographically and being pinged from the exact same location. Those burst translate in the game too making it unplayable. If the PoE gateway performed like my server, the game would be perfectly playable at 120ms with max bursts of 250 ( very acceptable ) Playing with 800+ms burst ( sometimes goes beyond 1200ms ) every 40 seconds is not acceptable and reflects that there is a problem on your side.

It looks like there is a problem related to the network where your server is located. It looks like some kind of bottleneck on some buffer that makes you stop receiving data for a while, thus spiking on latency. I will remove the possibility of being a problem of performance of the game client / server code, but It could be something at the network level, maybe the load balancers or the communication between the servers on the cluster ? A faulty switch interface ? Or it could be just that the network at that datacenter is poorly engineered. I obviously don´t own data as this is your project, so I can only guess.

But it is clear that you don´t need to be ingame to see the bursts problem, as we have demonstrated. It is a networking problem on your datacenter.

The lag bursts issue has been claimed in here more than one time, specially on the Singa server ....

I hope this is going to make change something, I will post more data soon.

Thank you for reading Brian.

I can only say that what chromafunk says is correct in my oppinion. Today I suffer so horribly from these "bursts" that u can't actually call them bursts anymore....its perma 1k-ping....