Right, like the sub-millisecond trading apps we have running on ESX at work in our production VMWare farm.
Again. It's fine. It works. The hypervisor does its job, and it does it well, as long as you do all the reasonable things required to keep your clock from having more than one source for its time. The #1 source of clock skew in a VMware environment is forgetting about the internal hypervisor's timing threads and relying on built-in NTP clients. By default, Windows has its own internal NTP client. You have to disable that. In a windows domain, by default you're doing NTP between the client and the domain controllers to keep Kerberos ticketing in check. So on your DC's, you have to do the same thing. In an ESX farm, the server hosts should have a single source for pulling time, which is why they build an NTP client and server into the host hypervisor. There is no frame-by-frame timing requirements by HLDS, or whatever it is you're talking about. Everything in HLDS is spammed via UDP which can't even guarantee delivery, much less make any time requirements. So again, even 100 tick CS:S servers are absolutely fine in VMWare provided you disable all other NTP sources within the guest OS, and you rely on the VMWare tools to do the job, since that's the process that syncs with the hypervisor. If you do that, you will not see any clock skew, even at 100% cpu load. ________________________________________ From: [email protected] [[email protected]] On Behalf Of Steven Hartland [[email protected]] Sent: Monday, February 09, 2009 5:05 AM To: Half-Life dedicated Win32 server mailing list Subject: Re: [hlds] VMware Advise lol @ NTP that's no good for accurate timing require on a frame by frame basis. Try running any code which will report time going backwards errors and you will see it happening, doesn't even have to be under load. Regards Steve ----- Original Message ----- From: "Karl Weckstrom" <[email protected]> > This isn't an issue under ESX or ESXI - at least if you enable NTP on the > host, enable vmware tools on the guest, and disable > built in NTP on the guests as well. > > It's part of VMW's best practices. Clock Skew under ESX/ESXI is a newbie > mistake. > > Also - instructions in vmware are not emulated. Virtualization is NOT > emulation. This is another common misconception that's > nowhere near true. > > TrashedGamers runs completely on ESXI and Openfiler, and it's fine. No clock > skew. No lag spikes. No timing issues. In fact, we > can pack things far more densely under ESXI than we can under Native 2008. ================================================ 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
