Going back to Windows/Linux, I run two Linux servers (12-slot 128 tick and 40 slot 64 tick) and this cvar made choke/loss issues disappear completely when it was raised to 128 000.
Like was mentioned earlier, many of us don't understand how or why, just that this was the golden solution for high load/traffic servers. I've been experimenting with: net_threaded_socket_burst_cap net_threaded_socket_recovery_rate net_threaded_socket_recovery_time But haven't found a good lot of values for these that produce the same effect. On Fri, Jul 11, 2014 at 6:29 AM, Winnie the Pooh <[email protected]> wrote: > Yes, but I only tested a few values when I did. Looking at net_showsplits > 1 it did seem to show a change but didn't seem to be noticeable in > gameplay. Personally, I have yet to experiment more with it. > > > net_splitrate was there along with net_splitrate_maxrate before as well. > > > "net_splitpacket_maxrate" - Max bytes per second when queueing splitpacket >> chunks >> "net_splitrate" - Number of fragments for a splitpacket that can be sent >> per frame > > > > On Thu, Jul 10, 2014 at 12:31 PM, Asher Baker <[email protected]> wrote: > >> Have you experimented with increasing net_splitrate since the change? >> >> >> ~~~~~ >> "Their heads are green, and their hands are blue, >> And they went to sea in a Sieve." - Edward Lear >> >> >> On Thu, Jul 10, 2014 at 8:27 PM, Winnie the Pooh <[email protected]> >> wrote: >> >>> Bruce, >>> >>> For me at least, I could see the choke and loss values in the in >>> net_graph on a live server drop as well as the the spikes on the bottom of >>> net_graph 5 go away when the value was increased. Of course, over all >>> smoothness of the server was noticed as well - movement and shooting seem >>> less "choppy". At the default rate it almost feels like friction is added >>> your movement slightly every few seconds. I'm not sure if that the kind of >>> description of symptoms you're looking for, lol. Maybe someone else can >>> describe it better. >>> >>> I'm not sure about now, but before this update if you had a live server >>> filled 24+ and changed the values while playing and looking at the >>> net_graph you couldn't miss the change. >>> >>> >>> On Thu, Jul 10, 2014 at 11:59 AM, Ejziponken - <[email protected]> >>> wrote: >>> >>>> 32 players 128tickrate = choke for all players >>>> >>>> And what about the other huger problems all servers have? >>>> >>>> I think there is still some issues with the servers and "sessions". >>>> Sometimes new players cant join a server until the server becomes empty or >>>> restarted. >>>> So lets say a server has 20/24 players. Nobody can join and at some >>>> point those 20 players will stop playing so the server will get empty and >>>> then new players can start joining again. >>>> >>>> And I think players trying to join get "could not find matching >>>> gametype" in the console. Or "session is full" >>>> >>>> Its really hard to catch this bug tho but I can see our servers being >>>> full and 30 min later empty. >>>> I can see it several times a day and I've talked with another server >>>> owner saying the same thing. >>>> >>>> ------------------------------ >>>> From: [email protected] >>>> To: [email protected] >>>> Date: Thu, 10 Jul 2014 18:08:39 +0000 >>>> Subject: Re: [Csgo_servers] net_splitpacket_maxrate >>>> >>>> >>>> Good to know. We’ll see what we can do. >>>> >>>> >>>> >>>> The change to reduce packet loss was to increase buffering on our >>>> network sockets. This only affected Windows because the buffers on Linux >>>> defaulted to a reasonable size. >>>> >>>> >>>> >>>> Can you explain more precisely what you mean by ‘choke’? What are the >>>> symptoms and when do they happen? Thanks. >>>> >>>> >>>> >>>> *From:* [email protected] [mailto: >>>> [email protected]] *On Behalf Of *Ejziponken >>>> - >>>> *Sent:* Thursday, July 10, 2014 10:28 AM >>>> *To:* [email protected] >>>> *Subject:* Re: [Csgo_servers] net_splitpacket_maxrate >>>> >>>> >>>> >>>> On linux I had to set it higher than 80k to get rid of the choke. >>>> >>>> >>>> >>>> >>>> ------------------------------ >>>> >>>> Date: Thu, 10 Jul 2014 10:02:05 -0700 >>>> From: [email protected] >>>> To: [email protected] >>>> Subject: Re: [Csgo_servers] net_splitpacket_maxrate >>>> >>>> Thanks! The more you know. >>>> >>>> >>>> >>>> Would re-exposing the variable and setting the default value or minimum >>>> value to 80000 be an options instead of blocking it altogether? Or do the >>>> changes made render the affects of that variable useless altogether? >>>> >>>> >>>> >>>> What does the change mean for Linux servers? Did they not suffer from >>>> the problem in the first place or? >>>> >>>> >>>> >>>> Thanks again. We appreciate all the information we can get. >>>> >>>> >>>> >>>> On Thu, Jul 10, 2014 at 9:54 AM, Bruce Dawson <[email protected]> >>>> wrote: >>>> >>>> I agree that the old default value of 15,000 was very low. Setting it >>>> to 128,000 would allow split packets to be transmitted much faster. >>>> >>>> >>>> >>>> However when net_splitpacket_maxrate was removed we switched to using >>>> the global data rate, which defaults to 80,000. So, if you were previously >>>> setting net_splitpacket_maxrate to 128,000 there should not be a dramatic >>>> difference now. >>>> >>>> >>>> >>>> The better we understand your current symptoms the better we will be >>>> able to understand the underlying issue and how best to address it. The >>>> change fixed some real packet loss problems for Windows servers so simply >>>> undoing the change is not a good option. >>>> >>>> >>>> >>>> BTW, the lack of mention of this change in the release notes was merely >>>> because we make too many small changes to list them all and this was >>>> assumed to be a small change. >>>> >>>> >>>> >>>> *From:* [email protected] [mailto: >>>> [email protected]] *On Behalf Of *Winnie the >>>> Pooh >>>> *Sent:* Thursday, July 10, 2014 9:32 AM >>>> *To:* [email protected] >>>> *Subject:* Re: [Csgo_servers] net_splitpacket_maxrate >>>> >>>> >>>> >>>> Bruce, >>>> >>>> >>>> >>>> The effect that most of us noticed and are now missing is that setting >>>> that cvar to a higher value eliminated some choke and loss especially on >>>> 128tick and higher populated server. The difference between the default >>>> value and max value was very noticeable on a fully populated server. >>>> >>>> >>>> >>>> Honestly, I'm not sure many of us know what the variable actually did >>>> behind the scenes to help but experimenting with it got us better results. >>>> >>>> >>>> >>>> On Thu, Jul 10, 2014 at 9:24 AM, Bruce Dawson <[email protected]> >>>> wrote: >>>> >>>> What effect are you trying to achieve? What was the advantage of having >>>> a separate rate for split packets? >>>> >>>> >>>> >>>> Historically the separate rate was to avoid packet loss but we >>>> increased the size of the UDP buffers on Windows to better resolve that >>>> problem. >>>> >>>> >>>> >>>> *From:* [email protected] [mailto: >>>> [email protected]] *On Behalf Of *Mohammed >>>> Khalik >>>> *Sent:* Thursday, July 10, 2014 7:04 AM >>>> *To:* [email protected] >>>> *Subject:* [Csgo_servers] net_splitpacket_maxrate >>>> >>>> >>>> >>>> net_splitpacket_maxrate 128000 >>>> >>>> >>>> >>>> Has anyone found a alternative cvar, or a workaround to achieve the >>>> same effect? >>>> >>>> >>>> >>>> >>>> >>>> >>>> _______________________________________________ >>>> Csgo_servers mailing list >>>> [email protected] >>>> https://list.valvesoftware.com/cgi-bin/mailman/listinfo/csgo_servers >>>> >>>> >>>> >>>> >>>> >>>> -- >>>> -Winnie the Pooh >>>> >>>> TimetoKill.net >>>> >>>> >>>> _______________________________________________ >>>> Csgo_servers mailing list >>>> [email protected] >>>> https://list.valvesoftware.com/cgi-bin/mailman/listinfo/csgo_servers >>>> >>>> >>>> >>>> >>>> >>>> -- >>>> -Winnie the Pooh >>>> >>>> TimetoKill.net >>>> >>>> >>>> _______________________________________________ Csgo_servers mailing >>>> list [email protected] >>>> https://list.valvesoftware.com/cgi-bin/mailman/listinfo/csgo_servers >>>> >>>> _______________________________________________ Csgo_servers mailing >>>> list [email protected] >>>> https://list.valvesoftware.com/cgi-bin/mailman/listinfo/csgo_servers >>>> >>>> _______________________________________________ >>>> Csgo_servers mailing list >>>> [email protected] >>>> https://list.valvesoftware.com/cgi-bin/mailman/listinfo/csgo_servers >>>> >>> >>> >>> >>> -- >>> -Winnie the Pooh >>> TimetoKill.net >>> >>> _______________________________________________ >>> Csgo_servers mailing list >>> [email protected] >>> https://list.valvesoftware.com/cgi-bin/mailman/listinfo/csgo_servers >>> >> >> >> _______________________________________________ >> Csgo_servers mailing list >> [email protected] >> https://list.valvesoftware.com/cgi-bin/mailman/listinfo/csgo_servers >> > > > > -- > -Winnie the Pooh > TimetoKill.net > > _______________________________________________ > Csgo_servers mailing list > [email protected] > https://list.valvesoftware.com/cgi-bin/mailman/listinfo/csgo_servers >
_______________________________________________ Csgo_servers mailing list [email protected] https://list.valvesoftware.com/cgi-bin/mailman/listinfo/csgo_servers
