[bf1942] BF2 server discussion

Steven Hartland killing at multiplay.co.uk
Sat Oct 23 22:49:19 EDT 2004


----- Original Message ----- 
From: "kama" <kama at pvp.se>

> While the 5.3 seems to be the new stablebranch, there are things to
> consider. with 5.3 the ULE will not be default as first anounced. it will
> still be the old scheduler (4BSD). This due to the fact that ULE is
> considered broken and the time fixing it will break the timeline
> increadably. In 5.3 there still will be a lot of slow code implemented,
> but when 5.3 is released, they will start tweaking the system. 5.5 will
> probably be the version where you will see what the new system is getting
> a boost in performance and all the experemental feature bits have fallen
> in to place...
This agrees with what I have seen for the sheduler ULE should not be used
for game servers are it definitely introduced bad lag as it stands.

> I have noticed that 4.10 is much more responsive than 5.2.1 under load. I
> have set up a system that handles a lot of trafic. the 4.10 could handle
> aprox twice the load. I am currently thinking of switching my 5.x
> gameservers to 4.10 due to how it is operating with load.

This should not be the case for game servers so not really an issue.
If your game servers are operating with high load they will be delivering
very poor performance.

> One other thing to consider is that when BF2 is released, FreeBSD is
> probably bumped to at least 5.4... I think the 5.3->5.4 cycle is going to
> be quite short. There are a lot of things that are ready but not
> implemented to the 5.3-branch.
> 
> One thing they have fixed is the nasty kernel spinlock bug. I have only
> encountered the bfvserver to work with the system in such way that this
> bug occures. hlds, srcds and ucc are not making the kernel panic. The bug
> happens during closing TCP-connections. And are the program closing the
> connection "wrongly", this spinlock kernel panic occures. This might be
> the BBO that are making the spinlock on the bfvserver. This server is now
> 4.10 and people are reporting that it is more responsive than before.

With all the bf servers we run under 5.2.1 we have not encountered this
ever so not sure how much of an issue it is. BBO is very dody in terms of
how it affects a bf server, so much so that we will not run it; so the issue may
lay more with that than anything else.

> Andreas: One thing that are annoying is the problem that you need to set
> the HZ quite low to have the bfX-server to opereate correctly. When I had
> it bumped to 1000, all crazy things happened. client-connections got
> closed the server started to get laggy.. when setting it down to 400 on a
> SMP system the server started to operate much smoother.. this is kind of
> backwards.. All the other gameservers get a bump when setting the HZ
> higher.

We run under a 200 HZ kernel and all seems to be good. Dramatically
changing the kernel HZ can certain affect the load presented by a server
we tested quite a few HZ values on many different games before settling
on 200 HZ as a good value.

    Steve / K


================================================
This e.mail is private and confidential between Multiplay (UK) Ltd. and the person or entity to whom it is addressed. In the event of misdirection, the recipient is prohibited from using, copying, printing or otherwise disseminating it or any information contained in it. 

In the event of misdirection, illegible or incomplete transmission please telephone (023) 8024 3137
or return the E.mail to postmaster at multiplay.co.uk.




More information about the Bf1942 mailing list