Dedicated server lag

[sneaky|vodka]

So ARK is not optimized.

There are tools like "-sm4" and foliage removal, but what about servers? I'm hosting a 50 people server on some of the best hardware, but ARK server doesn't use it all. In fact it uses like 3-5 Mbit of 1 Gbit bandwidth, barely any CPU and RAM.

Is there a way to tweak the server or maybe specific optimization for Windows Server 2012 that would allow it to use more bandwidth and lag less? An advice from the devs themselves?

Thank you.

[sneaky|vodka]

bump

[sneaky|vodka]

Bump

[Hades]

Maybe the lag is caused by combinations of latency and packet loss?

[sneaky|vodka]

I'm researching the forums right now and I see many people addressing an issue of low server framerate and unused CPU resources. I wonder if there is a way to help it.

[Exivus]

Wanting to know the same thing. The "server framerate" (tick rate) hovers between 4-6 at 50-70 people and goes down to 2.5 with lots of people, causing AI-controlled wild dinos to rubber band and lose their effectiveness completely. Can we get the devs to specifically address this for high-up server admins?

[vectorsky]

The biggest lag issue my server currently faces is when someone with a poor connection attempts to connect or actually does get in and plays for a while. Packet loss is a killer on the server atm.

I haven't found a way to mediate this at present other than identify problem players and limit their access which is not really a great solution but it keeps the server in a nice healthy state for those with decent systems and connections.

[S.jc9181]

the only major problem i have on my server lag is when an attacking dino clips and get stuck on the part its attacking the ping goes from a mear 45 to well over 200ms. Onces its dead everything is back to normal.

[sneaky|vodka]

I read the rubberbanding and lag are related to server framerate being so low with a lot of people on. In turn that is caused by the fact that server software doesn't use multiple cores, it only runs on the first core therefore not using all processor's resources.


@Drake, is there a way we can get at least a rough ETA as to when you guys will implement support of multiple cores on the server side and fix other issues mentioned above. Also if there is something in development for servers and it requires field testing, I'm totally willing to participate. Thank you.

[Exivus]

I would really like a specific update on this issue. This is the #1 thing preventing us from growing or supporting 127 concurrent players without incurring significant performance issues.

[sneaky|vodka]

Exactly.
I don't demand to fix it or anything, I understand it is pretty hard to restructure the code from one thread to multi thread and make sure it is all thread-safe.

Just wondering if this specifically is being worked on at the moment, if so what stage it is at. If not what are ETAs, because surely you guys [Wildcard] seem responsible and I know you have a plan.

[Hades]

From what I've seen, the server lag for any decently spec'ed machine... is caused by server frame rate (the game itself). This is largely independent of CPU and memory and IO (some people are running serious hardware out there) and as RightHand pointed out... there is diminishing returns throwing more and more hardware at the problem.

The Physx re-write is just off-loading server-side calculations to compatible clients... but since CPU isn't the issue for many of us it really only helps the low-end servers.

The fix is something the devs need to do. Solutions range from multi-threading parts of the game to putting certain calucations on a slower update cycle.... which is all dependant on what you can and cannot do with the unreal engine.

Anyway, the devs know this is important - I suppose all we can do is keep having friendly, thread bumping conversations to remind them :)

[sneaky|vodka]

Server framerate is not exactly what it sounds like. In reality it is the server's speed of calculating dinos movement and such. It depends on CPU very much. If you at load bring up by Core statistic you will notice that the Core 0 is very heavily used (maxed out), but the rest is like on halt. If the server was multi core this wouldn't happen and the server framerate would go up.

[Hades]

I think we are agreeing... in a round-about way heh :)

Yes, there is high utilization on cores.... but the solution isn't to throw CPU/memory/IOPS... and the root issue is the server frame rate (the game itself doing all those calcs on dino moves etc. as you say)

To fix server frame rate, you need to code the game better. So per my above post... writing it to multi-thread better or putting certain calcs on a slower update cycle are some examples of how that could be done.

One thing is for sure... this is all in the devs hands... and us throwing money at the problem isn't going to really help.

[sneaky|vodka]

Exactly. I hope this gets noticed.

[Exivus]

Devs, can we get a response on this? Very important to those of us that host your game.