MASSIVE desynchronized issue in-game
Issue is still happening. I got dc’s from a Tequatl raid due to lack of downstream bandwidth from the game servers.
Usually when I join a mass group event that has about 140 players, the game is not keeping me up in real-time, but slowly lags me behind by massive seconds against all the actions see on screen. Where I try to go or attack is not what others see on their side until 5 to 30 seconds later, or even higher. Thus, the actions between the server and the client gets desynchronized.
I’m sure your game has more bandwidth to provide since I can download updates up to 1 MB/seconds; yet when mass group event happens, you only provide me around 40 KB/sec, which is not enough when an area filled with players are blasting all kinds of attacks. Eventually all the data gets “backlogged” and I have to run away from the event and hope all that data “catches” up, or the backlog will overflow and disconnect me from the map. It often gets bad to the point that the game is not allowing me to tag enemies during an event when I’m up close to them.
Recently your September patch promises performance improvements, but since the patch went online yesterday, the connection lag from the game has gotten much worse; thus, the performance fix failed to meet expectations and making certain events much more unplayable.
I’m pretty sure something better can be done to prevent this data from “backlogging” on my game client. No games I’ve played ever lags me behind by 30 seconds or more.
There are two relevant bottlenecks where this could occur, and my guess is you are experiencing both at the same time.
First, if your client cannot keep up with the sequence of actions called for by the stream of packets from the server, and it gets far enough behind, you will DC. I chased an issue like this in DAoC a decade ago and finally realized that the problem was that the client side engine was being asked to do a heavyweight software call to instantiate an object on each {player, NPC} affected by a spell, and then de instantiate the object. This was why 200 people could stand around just fine, but once combat began the client performance collapsed. DAoC bought their client engine. I more recently observed similar symptoms in RIFT during its first year after release, and see the same in GW2, but have honored the Terms of Service and not run a debugger on either. I believe the client code should detect that it is getting behind processing traffic from the server and ruthlessly prioritize (meaning drop) ephemeral things like spell effects to make sure the keep-alive messages are processed in a timely way.
I also observe that if 140 players each have 40KB/s, that’s a bit over 50 megabits/sec, which is quite modest traffic for the server’s NIC to handle, but would by itself fill a fairly expensive uplink to NCsoft’s ISP there in Texas. If there is a device upstream of the server which limits the peak bandwidth each server can send to NCsoft’s ISP, and does so by dropping packets which exceed this limit (a process called “policing” in the network business), we would see this kind of desynchronization between server and client: if the server to client traffic is TCP, a dropped packet means wait 2 seconds and then retransmit everything following the drop. The telltale sign that this is occurring would be a fast motion catch up after combat ends. Just a thought…