The FBSD benchmarks I have only heard referred from former colleagues who were otherwise very much inclined to use FBSD over linux for hlds. The little experience I have had with optimizing and debugging FBSD hlds installations is a bit old, so it might not be the current state of affairs. But in these cases it had a noticeably higher cpu usage on similar hardware, although not critically so. 10-15% was the average.
As for the logging I did some tests on it some time ago, and found out that moving log-files to a ramdisk increased performance a bit - even on high-performance scsi drives. Same effect came from having map-files on ramdisks. It was noticeable but again not critical. It reduced map loading time by a second or so (estimate, as the uncertainty on that sort of stuff is huge owing to random processes running on the same server - other hlds's for example). So my conclusion was that 1) FBSD is minimally slower than linux for hlds use and 2) IO operations had a very small but noticeable impact on hlds performance. I'm sorry I don�t have any more specific benchmark results for you, but I did this for my own curiosity and not for publication. Also my tests were done about a year ago so lots of stuff may have changed since then. Morten -----Original Message----- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Steven Hartland Sent: 15. oktober 2004 12:30 To: [EMAIL PROTECTED] Subject: Re: [hlds_linux] Dual CPU Usage? ----- Original Message ----- From: "Morten Aar�e" <[EMAIL PROTECTED]> > FreeBSD is generally some 10-15 % slower for hlds use in my > experience. Interesting there Morten can we see your figures that led you to this conclusion? > Also please don't go so hard at the guy for making a valid point just > because you don't care to think it through. Although it is true that the > volume for the logfiles is very low the IO usage is quite high. Please > consider that every time _anything_ happens on either server the > harddrive needs to reposition it's heads and write it. In my experience > the buffering of logs is not really that efficient in hlds. Consider > further that while doing this an interrupt has been sent pausing > everything else for a few cycles (this is not entirely true, as the > chipset and processor of course have optimizations). Depending on the > level of caching etc on the harddrive and so on, the delay of such IO > operations can be noticeable. Again do you have an research to substantiate this? We've not seen / had any reports from issues here. 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 [EMAIL PROTECTED] _______________________________________________ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux _______________________________________________ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
