Lagspike problem workaround + Analysis
Hello,
Many people recently seem to complain about lagspike problems they experience in some maps. I have the same issue and I came up with a workaround and analysed the problem enough to come to a sophisticated answer. The last league I played was the harbinger league, with the same Internet provider, the same router and nearly no problems with lag whatsoever. This issue is new for me since I started playing in Beastiary league. Firstly: Of course I sought the problem on my end first. I analysed my MTR and the only packetloss I found was at the last *.networklayer.com hop so for me it was located in GGGs backend. Also there were no jumps in latency. I also contacted my provider and they checked the route to the server, which was ok. Also I use a high end gaming router with QoS configured for gaming and I made sure there was no other traffic interfering with PoE. I am also in contact with GGG tech support, but even tough i reported my findings to them in quite some detail they didn't reply to me with a statisfactory answer. The problem: You enter a Map/Instance and your Latency spikes to triple your RTT and if Lockstep mode is enabled the game freezes when then spikes occur. With more mobs on the screen the effect increases exponetially until the game nearly locks up totally. It can also happen that your Map is lagfree but if you enter a seperate Bossarena of a map it is lagged out. The fact that being in your hideout which is totally lagfree and entering a map which is totally lagged out and then transistioning to a boss area which is totally fine again made me curious, so I finally found a solution. The resolution: If I enter a map and I start noticing this laggyness I open a portal, go to my hideout and enter the map again. In 90% of the cases if I do this the lag and locking which was present in this map seconds ago is suddenly gone. The analysis: I am 99% sure that the problem is located in GGGs server-backend. The lagspikes seem to point to paketloss due to the way how lockstep mode works. With lockstep enabled the server confirms the packets with the client and if a packet doesn't arrive at the server he queries the client to resend the packet, and "locks" until the missing packet(s) arrive(s) and he can continue with all packets in order. So a packet which has to be resent has triple the RTT, hence the lagspikes. If 2 packets are dropped in a row you look at 6 times the RTT and so on. This also coincides with the problem getting exponentially worse the more mobs are on the screen. In such cases more traffic is transferred and the probability of more packets in a row being dropped increases, if more packets overall have to be resent in lockstep mode the freezing increases exponentially. For people who have predictive mode enabled the issue should result in massive desynch and teleporting of characters and mobs, but I did all my testing in lockstep mode as it is much easier to analyze such a problem on a reliable protocol. So why does this happen only in some maps, and why does reentering the map solve the problem in most cases? Now to answer that we have to think about how instances are generated in GGGs server-backend. I assume that GGGs server backend has virtualized servers which use some kind of load balancing and every time you enter your map and the map loads, it gets transferred to a server which has ressources available. If your map gets transferred to a server which is under too much load, packets get dropped as they cannot be processed in time and the lagspikes which many players experience occur, hence reentering the map and transfering the instance to another server has a chance of resolving the issue. So in my opinion GGG either has a problem due to their server-infrastructure just not being sufficient enough anymore to cope with the amout of playes at peak hours (because this is when the problems occur for me the most often) or there is a problem with how their load balancing works. My conclusion: I would really reccommend for GGG to seriously look into this issue and fix it ASAP, as so many PPL reporting similar problems with lagspikes is no coincidence anymore and you cannot just brush it off as problems at the players side (altough of course in some cases it is). GGG should really take those reports more seriously (and also the non GGG tech voluntaries in the tech forums who often reply to those threads as being a "user problem" too quickly even after more and more evidence of a problem in GGGs backend piles up). I really got very frustated with this but luckyly after checking everything myself I found this workaround. I would also be interested if this workaround works for other people aswell and I hope I can help some people with this post until GGG acknowledges that there is a problem and starts working on solving the issue. FYI: I mostly play on the Singapore server but while testing also encountered the same problem with the same resolution on some other servers (Australia, Japan for example). Last edited by Zoddo#3014 on Mar 25, 2018, 12:29:23 PM Last bumped on Apr 2, 2018, 3:20:53 AM
|
![]() |
Great analysis!
My problem is sometimes, more often then not, I'll hit a bad instance and within a few steps will freeze and can't even launch a portal to get out of the instance. The worst is when this happens in lab and you have no options. |
![]() |
" Heck, it doesn't even have to be a map. I was playing yesterday trying to make progress as usual, enter Kaom's Dream, and within a minute of entering the usual lag spike hits. I'm sitting there for 15 minutes watching as nothing happens, aside from the occasional 2 or 3 seconds of updates before going right back to lag, and the entire time of the lag spike I'm still considered "online" even though nothing's updating on my end (aside from particle effects) and the 6 second "heartbeat" should've long since kicked in. The only options that work at this point are exiting to login or exiting the game; "return to character selection" has absolutely no response. | |
With the term map i mean any playfield in general actually.
Last edited by Zoddo#3014 on Mar 25, 2018, 2:11:09 PM
|
![]() |
Yeah, same problem, and I resolve it the same way: Reenter the instance, which seems to fix it 99% of the time; there seems to be a window of time where changing istance will result in lagspike map, just need to wait 30 sec and do the above to fix it.
Realy annoying, especialy when u open a map and u got limited portal, or even worse, when u got Zana mission and it's only 1 portal, or u just finished to defeat Izaro and there is no way to change istance untill u finish the sector... |
![]() |
Have same problem,since last patch (3.2.1b). Sadly the workaround never works.
|
![]() |
I've been having this problem too. It's cost me at least 80c because I kept dying on forge maps. I'd lag, get stun locked, die, repeat. It also got my hero killed in a breach. The lag spikes to 100-200ms, which would be fine except that it locks the game up for a half second or second. at which point it seems to have canceled all my inputs.
At first I thought it was a graphics issue since I am running a 6950 and mob spawns and particle effects can cause lag. But I noticed this was only happening when the lat would spike significantly. This is a relatively new issue for me as well. If anyone has any advice or workaround that'd be neat but otherwise I'm not sure what to do. |
![]() |
I just described a workaround, maybe you should read it :) . Next time you enter a lagged out map/playfield try going to the hideout and enter the map again.
|
![]() |
" My hat is off to you. I trying this during my play session today. It works like a charm. Thank you and may I wish many exalts to fall for you! Keep on Truckin'. |
![]() |
" Glad I could help. |
![]() |