Limelight Forums

Full Version: Spawn optimisation feedback
You're currently viewing a stripped down version of our content. View the full version with proper formatting.
Pages: 1 2
I just made some optimisations in some of the player networking done on spawn (both initial spawn and spawn on death/change job). Things seem a lot faster on beta but this doesn't always work as well on live. Also, what works great with a few players may still be very slow with any more. 

Would you please be able to provide feedback on if you've seen any improvement on initial spawn on load since today, the 23rd, or general spawns (jobs/death?). It will help me to determine what any next steps here may be. Thanks!
I'm on now and the spawning in feels quicker but there is only 6 people on.
I loaded into the server pretty fast today (around 15-20 seconds). But there were only 4 players online.
I believe this has casued some more crashings, today Ive crashed around 4 times and usually I do not crash one single time.
I have noticed my load time fell from 15 minutes to 5 minutes when server is peaking. And the fact I am timing out less when trying to join with over 40 people on, from 4-7 timeouts to 1-3 timeouts.
Has caused me great joy, loading into the server about 70% quicker than before. Great job, .
I just noticed that I crash more on initial spawn than I have done before.
Thanks for the responses. There weren't too many so the results may be a little inaccurate, but hopefully this is a decent representation of what people have seen. I've closed the poll now as recent further optimisations have been made that could render these results invalid.

So from what the results tell me and I've managed to gather from server statistics, this update was fairly significant in reducing player networking overhead and spawn/load times. The additional amount of networked variables was quite severe for the system, and the code that would cycle through them and make changes on every spawn was very poorly optimised. With that all being removed and some small, optimised replacements added to manage the weapon mods system, the operations performed on spawn are significantly fewer. It makes sense that spawns and loads are improved.

It doesn't seem to have improved for everyone, though. This could be a consequence of them testing mainly on high population servers, or experienced their own unique issues elsewhere (F1 menu, their own hardware or internet connections, etc.) which we will try to tackle to improve spawn and load times for everyone. 

This is also just independent observation so may be wrong, but ping does seem to be reduced by a further 10-15% on heavily populated servers, which would make sense given the 15-20% reduction in networking overhead in this update. 

I think my next steps in this area will be to either further reduce player networking if possible. Does everything that is networked need to be? I'm not sure. Certainly one or two more things can be cut, but perhaps we can reduce a lot more. Weapon mods are no longer networked at all - I don't think other things need to be either.

I'm also going to investigate either disabling the F1 menu on spawn/load, or having a derma page load as opposed to a html page. In our future work on the HUD and F1 menu, clientside web files will be loaded, hopefully further increasing performance. Noble and Mathias are doing testing on that right now.

Again, many thanks to those who voted. Please feel free to ask any further questions, otherwise I'll close this thread shortly.
Speaking to what Faustie mentioned it has become apparent that loading the webpages from the webserver onto our F1 menu does cause some significant hanging and the new F1 will definitely be built to avoid this completely. This should further decrease load times once in the server and hopefully prevent F1 menu crashes that we've seen from time to time.
The other thing that is happening a lot now is freezing while on foot or driving
any updates coming to fix it?
(Feb 28, 2018, 06:07 PM)Gamerpro2266 Wrote: [ -> ]The other thing that is happening a lot now is freezing while on foot or driving
any updates coming to fix it?

We'd need more specific information in order to address and fix the issue. The more information/error codes/etc. that we can get the easier it is for us to diagnose the issue.

Let alone it may be a hardware issue
(Feb 28, 2018, 06:07 PM)Gamerpro2266 Wrote: [ -> ]The other thing that is happening a lot now is freezing while on foot or driving
any updates coming to fix it?

As Noble said, more information please. Is this at low or high populations? Does there seem to be any relation to nearby Photon lights, or large amounts of props? Is your fps and ping otherwise stable?

Also please check your disk health and I/O when this occurs. This kind of freezing is very often indicative of failing or slow disk hardware.
I do believe it's server side because when the game freezes you hear a couple of explosion from where car cashed into walls and when I asked someone else they said they got it.

The freezing started around when doctor internet remove atmos 6328 and clean up location code for client 6329

I'll keep you updated.
(Mar 1, 2018, 12:20 PM)Gamerpro2266 Wrote: [ -> ]I do believe it's server side because when the game freezes you hear a couple of explosion from where car cashed into walls and when I asked someone else they said they got it.

The freezing started around when doctor internet remove atmos 6328 and clean up location code for client 6329

I'll keep you updated.

Those 2 updates were not live when you reported that.

ForceGhost

(Feb 28, 2018, 06:07 PM)Gamerpro2266 Wrote: [ -> ]The other thing that is happening a lot now is freezing while on foot or driving
any updates coming to fix it?

I thought I was alone with this one! This is the number 1 reason I slam my cars into a wall.

  • It happens at low pop and high pop. Sometimes 1-2 seconds sometimes 4-5 seconds.
  • I've noticed it more then driving but occasionally happens on foot too.
  • Photon doesn't seem to have an effect on it.
  • Ping is always below 50ms.
  • FPS is usually stable but just suddenly tanks at seemingly random intervals.
  • No errors in console that I recall.
  • Diskhealth and I/O are fine.
Pages: 1 2