Lagging Servers. IMPORTANT FOR SERVER OWNERS

I have seen a massive surge in servers trying to allow more players than they are suitable for, and with a higher vehicle limit than they can handle. this leads to slow, laggy gameplay which is not enjoyable for the players. No player wishes to play on a server with 8 extra players than the next server if the gameplay isn't fun.

It does not matter if you can hold 64 players on counterstrike, empires must work with the difficult vehicle code in the source engine and this will create lag... not to mention the far larger maps, dozens of buildings on each team among a hundren other things that many other source mods won't expect of the server.

The end result is that people think that empires is laggy.


so a plea from a player to all server owners: check that your servers are not lagging, and if they are, reduce the amount of maximum players, and reduce the max amount of vehicles purchasable per team by using this server command "emp_sv_max_vehicles" (changable in server.cfg in empires/cfg folder). for most servers, 8 is a good number of vehicles per team to ensure no lag, and for larger servers with 48 players, IF you can handle that which i have yet to see a server do, even 6 - 7.

 

This is very true, but people really haven't been able to test out 48 player maps with the new version until now.

I expect servers to balance out over time, with the admins dishing out for upgraded hardware or lowering their player slots.

 

And for players experiencing laggy gameplay, try moving to the official server. That will always be balanced with settings promoting the best possible gameplay experience.

 

Well I have 100mb connection using a Quad Core Xeon 4GB Ram and I didn't modify the default server.cfg I just was asked to experiment with the 48 slot and I noticed the CPU sometimes gets a bit high but normally it's running around half usage or below but the network for the Empires server is almost nearing 100% usage which isn't good...

I was hoping to see if anyone here was sharing their 48 slot server settings to maybe help me reconfigure my server otherwise after a few days I am reducing the slots to around 32 again...

I will lower the vehicle count but I really think it's more to do with the maxrate setting at 30000 and I know I need to keep it there but I think because that is so high it really taxes servers hosting the mod...

I am not complaining what so ever just stating the obvious and I will do what I can to minimize lag but seriously there are other issues at play here like some members of my clan lag really bad just because they set their graphics up to high and first they blamed the server when nobody was even lagging but I think it's their Ego getting in the way and seriously not all server owners are going to minimize servers to accommodate old client hardware and poor or far away INTERNET connections as it would ruin the game play for those that have those beefed up machines and are on Fiber Optics!

Anyways that's just my two cents and I am sleepy!:p

 

Please use a low tick-rate (33 is fine) to prevent massive network lag. 100-tick servers with 32 players will use up appalling amounts of bandwidth and probably choke the server near-instantaneously.

Graphics settings do not cause higher latency, they cause lower FPS :p

 

yeah I have a 8800GTS and I can seriously recommend not going for the highest settings as it still lags the hell out of me at times..

 

hey caelo, easy on mocking the guy, he is providing an immense server for the game.

Thanks for that, CoZmicShReddeR

 

No offense taken I always am used to people trying to correct me!

If you had read what I posted it wasn't meant to make it sound like I consider everyone was stupid it was meant there are actually people out there that blame the mod or the server if they have a crappy graphics card or computer...

I been hosting game servers since 1997 and I have seen and heard it all...

 

Hey guys, Term from FragArena, that other 48 slot server.



We're currently running our Empires server on our primary server which has:
Single Processor Quad Core Xeon 5335 - 2.00GHz (Clovertown) - 1 x 8MB cache
4 GB FB-DIMM Registered 533/667
250 GB SATA II WDHD x 2
100 MBits I/O

Obviously our server is a monster. I've been watching the server for some time, and I have a few questions I would like to ask Krenzo about the whole entire thing. But anyways, our server is running at 33 Tick (Asplood from anything above it I'm sure) with fps_max at about 100 (anything above with cause asplood!). I've noticed a few things, but mainly when the server is full the OS reports around an 18% CPU usage on the server with very little PM actually in use (compared to what we have).

The server really never lags unless FPS gets low because of the vehicle count near 32 (our max) on both teams. At any given time with a large amount of vehicles, it can drop as low as 12 FPS, and I would imagine it would be faster because of the processor clock speed, but that's not the case apparently. With that said, the server really has no lag unless those vehicles get too high. Sometimes people complain of lag, but it's either when the FPS to the server is starting to get a bit too low or if their PC can't handle all of the vehicles and 48 people in the server with everything going on.

With that said, my question to Krenzo or another dev. Basically, I was told by someone that SRCDS does NOT support Multi-Core threading and basically runs from the first core/processor available. I personally don't believe it because of Valve implementing of 64 man servers for CS:S and other source mods. What logic would it be to allow so many people in one server, yet only support one core/processor for the whole entire task? Although there is some evidence that says otherwise, at this point I don't know. SRCDS when full for our Empires server reports 80% CPU usage, yet the OS reports a very low CPU usage with around 10%-20% actually being used. My question to you is: does the server ACTUALLY support Multi-Core functions? It looks like it does, except it calculates CPU using an old method. And the cores all stay the same the whole entire time rather than one going rouge and having a higher usage. But, from looking at the source and such, I'm sure you all may know the answer. Reliable people should be questioned, right?

And a few responses to finish it off:

Tick does not lower. If you specify what you want the server to run at, it wont change. It will stay constant, even if that means overloading the whole entire machine and crashing the server.

Exactly. The difference between 33 - 66/100 tick is the difference between night and day. But, it really only is recommended to use 100 Tick on servers with 24 and below players because of the amount of stress it will put on the server. Competition servers are the majority of 100 Tick servers. Otherwise, there really is no point. 66 Tick is also very good, and should be used for servers using UP to 32 players. You'll only notice a slight difference. Anything beyond 32 players with 66 tick is just redundant and asking for lag. 33 tick is great for servers with a whole bunch of people. It doesn't cause lag and it uses just the right amount of resources to make sure the whole entire experience is nice and smooth.


Anyways, long enough post here. I'll leave it alone. Thanks guys! Awesome mod! Drop by in our server and give us a shout some time!

 

sweet jesus, am i reading that right? 32 tanks... PER TEAM?

Dear god, I'm playing that server.

 

Multi-core processors won't be supported until the Orange Box SDK update is released, and then Krenzo will have to merge the code and all that crap.

 

Awesome. Although it makes me wonder why the OS reports something totally different than SRCDS when it comes to CPU usage.

 

I'm not sure but my guess is that it is using 80% of the first core, and 80/4 is 20, which is what the OS is reporting.

 

I assumed that also when I first saw it, but I checked all four cores and all were sitting around the same place in the 10% area where the OS actually reported.

 

Well I set my server to use just one core and it does just that not sure what you mean about it using the first core I like to put my servers on different cores as I have four to use...

I am glad to see some success with your settings as I am a noob and vary slow at getting this stuff right and usually ends up with a vary successful server launch then ends up burning in hell after a few days!

I hope to get all this right and maybe be a bit more successful like any other well loved server host on the Sourcemod engine...

And I am glad to hear that Empires is or might go Orange Box sounds awesome and sounds like a load of work as well..

 

Dude 10-20% "from all cores" may be 40-80% on 1 core. SRCDS uses only 1 core. Orangebox uses separate threads for physics and main server code so it will probably use at least 2 cores/CPUs. But it isn't released yet and multi core servers only allow you to make like 4 game servers on one host. You won't get any speed up from moving from single core to multi core. BTW which OS is reporting 20%? Windows or Linux?

Source games require a lot of bandwidth and Empires has many more entities in the game than CS:S. It needs 30kB/s per client. But I can be wrong - I didn't try to play with lower rate lately - in 2.0. BUT do you try to say that Empires server with 48 slots uses up 100Mbit/s? That would be ridiculous. If so then Krenzo would like to hear about it I think.

There are other server variables that matter: for example cvars that changes rate of updating the minimap (emp_sv_player_minimap_interval and emp_sv_vehicle_minimap_interval).

 

No. As I have said in another post, I was checking each individual core while all FOUR were sitting at about 10%. While this makes sense, SRCDS was reporting 80% usage. Sound proportional? If you have eyes, it's obviously not. It is not only coming from one core you know.

Are you kidding me? This is so wrong it makes me chuckle. Not so long ago, I was hosting 7 servers, FULL may I add on this box with said affinity set to all four cores. No lag, servers FPS didn't drop below 300, and CPU usage was all below 30%. If that isn't an upgrade from the thing you call Hyperthreading with P4's, then I don't know what is.

If it isn't an upgrade, I ask you to host a 48 man Empires server with a P4 with a possible 2 GB's or RAM just for kicks. Since multi-cores really don't make any difference at all.

This is so untrue it's... wow. Multi-cores are a HUGE step up beyond single core processors in many ways. Sure, not in this case if what you're saying is actually true, but in many other cases with other games that do support multi-core processing and other server applications that are advanced enough to do the same.

Windows is reporting a 20% usage and a 10% on each individual core. It wouldn't make a difference with Linux proc check if that's what you're getting at though.

If any individual server was using 100Mbit's at max, that would be a huge problem. Fortunatly, it's not and never will be. When the Empires server is full, it will use around 5-6 MBit's P/S I/O in any case.

This doesn't answer anything. The issue at hand is not bandwidth since every server hosting a server seems to be doing fine. Bandwidth SVar's wont do anything.

And for others who are confused with cores and processors, I'll break it down since a few people are a bit confused. Cores are ADDITIONS to a processor using the same processor cache and bus interface. Cores are not individual processors! They are actually PART of a processor. Hints multiple cores, one processor. Dual processors are a totally different thing.

 

You won't get any speed up from switching from single core to multi core - it's obviously right for srcds. I'm not an idiot I know that some application might scale. SRCDS DOES NOT SCALE.

10% from all cores - Windows might be wrong.

Your response to:
"But it isn't released yet and multi core servers only allow you to make like 4 game servers on one host. "
I don't know what your' talking about. Why the hell do you compare HT with multi core? I said that single threaded (or mostly single threaded) apps won't get any speed up from multi cores.
I was generally talking to Cozmicshredder btw if you didn't notice.

 

The only source game that runs multithreaded is Team Fortress 2 right now. Counterstrike can handle 64 players with 100 tick because there are no physics and levels are a lot smaller. This leads to less cpu usage from both networking and physics. In the Orange Box engine, both networking and physics run in their own threads. That means that you will instantly get better performance once we move to Orange Box.