Improved stability and dedicated server - Discussion

While that is a problem, trust me I have seen it with my campaign work, its not really related to the topic of stability of servers and netcode I dont think.

Its the package as a whole. I would think something new that causes a crash in SP or MP by definition effects stability of MP. If anything its just another aspect that someone who hosts an MP server has to worry about or possibly debug in their code. At best its an annoyance and is viewed by them as the game being unstable or unreliable. That was my point.

Do you run SSD drives?

No, mechanical HDs

One of the issues with loading into MP missions is that the game has issues "finding" files. I know this as I've checked my log immediately afterward. It'll struggle to find certain files which will cause the program to hang. If it hangs too long in MP, the game will freeze or crash. Putting the game on 2 SSDs in raid 0 "fixed" my problem by giving the game greater speed to locate the files it struggles to find (so that it doesn't hang as long), but this shouldn't be the solution. game badly needs to optimized.

SSD drives make a huge difference, I have installs on both, the difference in loading SP and MP are pretty big.

I need a "donation" :megalol:

I am sure maybe in some cases, but I think Grimes was more talking about mission breaking events, but maybe I read it wrong.

It's both. Changes in behaviour break triggers and timings and alter how units react to scripting, which causes its own set of issues from one day to the next (ask anyone who has tried to build a mission, be it in MP or SP using troup transportation), but some changes will significantly alter how those behaviours are being processed.

AI is an immense resource hog, which is hardly a surprise, but that means that anything that alters where when and how that processing happens can also break missions — in particular in an MP settings — in a way that can actually cause more problems than just being outright broken.

If it's broken, you can change the methodology by using different triggers, spawns, unit/AI activations and so on to replicate some outcome that should have happened automatically had that now-broken scripting or trigger logic worked as it should. If it processes differently, you're just screwed because there is no way around that. The mission may still work… of sorts, but be rendered completely unplayable without a complete rebuild because there is no clean way of creating the effect you want without the new processing murdering the server CPU.

Eg. a change that makes AI units more aggressive in how they deploy certain weapons will suddenly mean that a well balanced mission is not just no longer balanced, because it was tweaked for a different behaviour, but also that it becomes almost unplayable because all AI units go macross-mad with missile spam (probably the worst thing in terms of creating server load per unit deployed). Of the two, the rebalancing is a pain, but still by far the easier bit to rebuild and cope with as, but the CPU-eating behaviour does not go away.

 

As for stability I have had very few issues with that and that has SHOCKED me because for the last few months I have been playing DCS Multiplayer in probably one of the most remote places DCS has been played in... a Tiny island in the Caribbean with less than 200 people on it at any one time called Little Cayman. The internet here is horrendous and yet I can still play DCS multiplayer with no issues or problems over a wifi connection, both as a client and as a host, especially in the small private servers that we put up in the multiplayer group r/WingmanFinder.

 

The fact that I could play and record us flying and having a good time while in this very remote area completely shocked me. This experience has led me to believe that a lot of players are trying to connect to public servers which are way over crowded with clients and objects. As well as players maybe not having machines that are up to spec for the heavy load of DCS multiplayer (of course who can blame them computer components are EXPENSIVE!)

This is pretty good information though. It means that your "horrendous" internet still being able to host multiplayer means that it's unlikely network problems but rather simulation problems (either introduced by the mission maker or inefficiencies in the game code).

I'm interested though.. How many people were you hosting? After about 10-15 people the upload required starts getting pretty unreasonable for average personal home internet, and starts getting into business / hardcore upload speeds. So having some numbers on the horrendous internet would be useful!

But yeah good info. I've been able to run missions like On Station and whatnot from my house, and these typically last a few hours (I guess that's what you meant by large?), but anything more persistent and things start getting iffy.

And thanks Spud, I enjoy your content as well. Keep it up!

Thanks for the support NineLine! My videos of the very causal group of guys having a blast together has definitely brought new people into the multiplayer fray and the way we have set things up people have reported very few issues even with systems that are not tippy-top of the line in terms of hardware.

If you ever want to join us for a flight check the Wingmanfinder discord we usually play friday and saturday night!

This is pretty good information though. It means that your "horrendous" internet still being able to host multiplayer means that it's unlikely network problems but rather simulation problems (either introduced by the mission maker or inefficiencies in the game code).

 

I'm interested though.. How many people were you hosting? After about 10-15 people the upload required starts getting pretty unreasonable for average personal home internet, and starts getting into business / hardcore upload speeds. So having some numbers on the horrendous internet would be useful!

 

But yeah good info. I've been able to run missions like On Station and whatnot from my house, and these typically last a few hours (I guess that's what you meant by large?), but anything more persistent and things start getting iffy.

We usually have an average Pilot count of between 8 and 16 guys and so far things have been great! (Knock on Wood) I think its the massive object counts on a lot of the big public servers that kills people's performance...

This is also paired with the fact that most large public servers are on the Caucasus Map and I have pretty significant dips in performance on that map due to the massive tree count even on free flights in single player where its only me and a wingman and that is with an i7-7700K, GTX 1080 and 32GB of RAM

This is why I LOVE the Persian Gulf Map the performance on that map is better than any other map we have at the moment in both single and multiplayer.

While our netcode certainly needs improvement, and the addition of dedicated server will certainly will help, this is also a big factor we believe (NOT the only one though).

 

Thanks,

Matt

A lot of the large public servers are also running on the Caucasus Map, whose large tree count also puts another strain on people's systems besides the large number of objects placed in the mission.

This is why I love the Persian Gulf Map the performance on that map is fantastic! And the geographic setup of the map is perfect for multiplayer with the two "sides" separated by a body of water making navigation very simple for most players and giving some recognizable separation in PvP.

I think its the massive object counts on a lot of the big public servers that kills people's performance...

This is a tricky issue, we run into problems at around 500 usually. The biggest problem is that it seems like wrecks count as units, and with no way to clear wrecks, we're kinda screwed there, as is any mission eventually that tries to spawn in units in a dynamic manner.

Any decent training environment in the NTTR basically requires a "massive object count".

Is it possible to have servers that will limit the player count? Obviously optimizations need to happen for large player count say 8,16, 32, 64.

I'm sure people are already restricting it to a number but then this would also mean using less objects, maybe a spawn system too rather than everyone selecting a plane and sometimes end up spawning on top of each other.

I definitely agree with the people talking about servers with massive unit counts and tons of complex scripts firing left, right, and center. However, I don't believe making MP missions less complex is really the answer. Dedicated servers alone would go leaps and bounds to solving the issues of mission complexity.

I think the problem here is that MP mission devs have hit the limit on mission making, and it wouldn't be pulling teeth to fix a few things to raise that limit.

I also wish ED would fold some of the more popular and helpful scripts into the engine itself (though that's a decision for the developers of those scripts). I think that would go quite a ways to making things a little easier on everyone.

Side note on the issue of SSDs: I'm just saying, Samsung on Amazon is selling 500GB M.2 SSDs for $130 USD. If you think your standard SSD is fast, this thing is warpspeed. I picked one up and I promise you from my experience, this thing will make an insane difference. I do not experience a lot of the loading problems for MP that others have since moving DCS to my M.2 drive. I get a 'connection timed out' error every now and then, but it's quite rare, and never more than once.

No, mechanical HDs

That sucks, HDDs and large DCS missions do not play well together I know how you feel - one workaround is to load the mission file of the server you're trying to enter in the mission editor first so it is pre-cached and then try to connect.

I'm getting the sense from Wags' and Nineline's comments that you partly see the problem being the way the bigger server's missions are set up - due to too many units or scripting or whatever. People play on DDCS, 104th and Blue Flag because they enjoy playing PVP across the entire maps with enough players for it to feel alive and enough ground units to make it interesting. I'm getting the feeling that whatever you're about to announce isn't intended to fix the issues those servers are facing but is designed to replace those servers with smaller, simpler missions that are easier for you to stabilise.

Is the dedicated server (or official server, whatever you're calling it) something DDCS would be able to use with its current mission?

This is a tricky issue, we run into problems at around 500 usually. The biggest problem is that it seems like wrecks count as units, and with no way to clear wrecks, we're kinda screwed there, as is any mission eventually that tries to spawn in units in a dynamic manner.

I think its more that the object at a scripting level is still there and is never removed. It applies to objects that don't create any wreckage or debris like ships or aircraft crashed into water. Thats a major scripting behavior that changed from 1.2.x to 1.5. Bug report is still open on it, so don't know if it was an unintended feature or not.

Just in the interim until we release the dedicated server and some other improvements to improve stability. The more I looked into the composition of some of the online missions, the more I see that many are far too overloaded. Once we get the DS out and some other improvements, it's my expectation that DCS can handle such huge missions and perhaps even larger.

The question is, then, how do we create the kind of battlefield dynamics that everyone wants without overloading the map? Someone mentioned a lack of best practices a couple of posts back, and this would definitely be an area where some good practices, or even hard-coded methods, would help in that case. As would some improved clean-up mechanisms — if nothing else than to allow for longer-term runs in general.

Also, I'd imagine that you'll find the PG map to be the most stable and the Caucasus map the least.

Not to start that discussion, but there is a very clear reason why everyone — including the server owners — run Caucasus as opposed to any of the terrain modules… ;)

I dont think the map specifically, but underlying parts of the engine did, as can be seen in things like deferred shading and such. So again, the demands and specs did increase when we went from 2.0 to 2.5.

Okay...but, it wasn’t warping like crazy until a few version back and now it’s almost constant.