I can concur this also. When L4D demo was released, I was running 8 forked servers on a CentOS 5 VM under ESX. I could've ran more, but bandwidth usage was pissing off the people who pay that bill. I run my own personal srcds server under ESXi here in hawaii. Yes, HLSW DOES in fact report lag spikes - I haven't been able to track down why, but what really matters is performance IN GAME. No Lag spikes, AT ALL.
No issues with timing or anything have popped up either. Granted, I've yet to try running any game servers under windows in a VM - I do all my personal game serving under linux - so I don't know if that makes a difference or not. I've ran a BF2 server in a VM, Quake 4 server, COD4, and an ET:QW server. No problems. I know that's not even close to being a comprehensive list of games, but they are known for requiring some hardware to push them .... FWIW anyways. /me goes back to complaing about hlds / hlds_l mailing list shutdowns ... Mauirixxx PS: Karl, I actually found your instructions on setting up ESXi with FreeNAS and iSCSI last week when I tackled that very subject here at work - it was a very interesting read. I really think you can expand on it that though, if you ever get time :P -----Original Message----- From: [email protected] [mailto:[email protected]] On Behalf Of Karl Weckstrom Sent: Monday, February 09, 2009 10:34 AM To: Half-Life dedicated Win32 server mailing list Subject: Re: [hlds] VMware Advise Yeah, there's very different behaviors between the two major types of hypervisors (type 1 - or "bare metal", and type 2, which requires a host). On a bare metal hypervisor, you're guaranteed high efficiency because the hypervisor has direct access to hardware. ESX and ESXI are examples of this. VMWare Server (formerly GSX), VMWare Workstation, etc - those use type 2 hypervisors. In this kind of scenario you *could* see problems with time-sensitive apps, but how bad these effects will be depends on the host OS you're using. Linux and Windows are both pretty decent here. In fact, the only time I've seen any kinds of problems with Type 2 hypervisors is when you do novel things like run VMWare Server on BSD with Linux Binary Compatibility. There's a lot of reasons to virtualize... Way too many to list here. But I can say with 100% confidence that SRCDS is completely well-behaved in ESX/ESXI under Linux, Server 2003 and Server 2008 guests (behaves on Server 2008 best, since MS is obviously aware that people are going to virtualize it whether MS supports it or not). -----Original Message----- From: [email protected] [mailto:[email protected]] On Behalf Of Steven Hartland Sent: Monday, February 09, 2009 2:14 PM To: Half-Life dedicated Win32 server mailing list Subject: Re: [hlds] VMware Advise My must say I haven't tried SRC itself but have tried others and they lag badly due, which I assume is due to use of comparison in timestamps and / or inaccurate sleep periods. With regards trading apps do they really require events to happen with ms accuracy? I cant picture a scenario for FD which would cause an issue if it happened +-a few ms. In contrast I would personally be very surprised if there was any game server that wouldn't see at least some detrimental effect from having inaccurate timing. You would only have to be doing any sort of time difference calculations in a simulation for an clock frequency variations / sleep variations to show noticeable issues. Its very similar to how game servers present issues when run on hardware with any sort of power saving, that makes use of variable clock frequencies, enabled. Would be great if some from Valve could comment on the potential issues for timestamps going backwards or inaccurate sleep intervals on a regular basis would be for HLDS / SRC based games? Regards Steve ----- Original Message ----- From: "Karl Weckstrom" <[email protected]> To: "Half-Life dedicated Win32 server mailing list" <[email protected]> Sent: Monday, February 09, 2009 6:12 PM Subject: Re: [hlds] VMware Advise > I'm well aware of the doc. I had to know it backwards and forwards for the VCDX. > > Trading apps care because of timed trade executions that have to happen at very specific times/intervals or when triggered by a > market event. > > What you're talking about Batch processing. That's not what capital markets trading apps do. > > HLDS/SRCDS is not a time sensitive app, meaning it does not require sub-second accuracy (much less single-digit ms accuracy), > which makes it an excellent candidate for virtualization. You won't see any problems with HLDS/SRCDS under ESX/ESXI. Apparent > time is more than enough for the app since the app seems to only care about time for one thing. Timestamps on logs :) ================================================ 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 +44 845 868 1337 or return the E.mail to [email protected]. _______________________________________________ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds _______________________________________________ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds _______________________________________________ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds
