i'm curious, what is the metrics used for the "s*" description of the Linux network stack? and are they referring to v2.4 or v2.6 ? :)
Some of you can refer to the Internet2 Land Speed Record as it is currently being dominated by Linux v2.6.x (a single entry from NetBSD is also present on the list). http://www.internet2.edu/lsr/history.html or to this paper from Passive & Active Measurement Proceedings of April 2005 : Measured Comparitive Performance of TCP Stacks by Sam Jansen, Anthony McGregor http://www.pamconf.org/2005/PDF/34310332.pdf .... TCP performance during 5% bidirectional loss --------------------------------------------------------------------------- OS Min Mean Max SD Linux 2.6.10 164.38 213.98 287.67 22.75 Linux 2.4.27 153.82 207.42 248.70 22.86 FreeBSD 5.3 136.77 176.20 225.01 17.11 FreeBSD 5.2.1 128.74 162.81 219.01 19.56 OpenBSD 3.5 63.84 117.98 166.82 22.11 ..... * measured in terms of goodput (the higher the value, the faster the performance). http://en.wikipedia.org/wiki/Goodput Summary .... During bidirectional random loss, the Linux TCP stack is able to obtain the most goodput by quite a long way. .... i think the FreeBSD team was also aware of the result, that in July of the same year (2005), made this annoucement: http://lists.freebsd.org/pipermail/freebsd-announce/2005-July/001012.html .... " The TCP code now needs a general overhaul, streamlining and cleanup to make it easily comprehensible, maintainable and extensible again. In addition there are many little optimi- zations that can be done during such an operation, propelling FreeBSD back at the top of the best performing TCP/IP stacks again, a position it has held for the longest time in the 90's. " .... or was "uneasy" about the stack as early as 2001 ? :) http://www.ddj.com/dept/linux/184404914 .... "sometimes one steals ideas. Linux, for instance, stole part of the BSD networking stack. [Pauses.] All of it. --Warner Losh (FreeBSD) .... here's another early Linux v2.6.0 & *BSD connect / http request latency benchmark (2003) : http://bulk.fefe.de/scalability/ despite of all these data, i think there's no point in arguing since the latest data may not hold for a long time. and who knows, redmond might catch-up... (fyi, they also have entries on the Internet2 Land Speed Record and they have vast resources). =D
On 2/6/07, Cocoy Dayao <[EMAIL PROTECTED]> wrote: >To: "Philippine Linux Users' Group (PLUG) Technical Discussion > List" <[email protected]> > Sent: Tuesday, February 06, 2007 7:50 PM > Subject: Re: [plug] [OT] NetBSD 3.1 LiveCD r o c k s ! > > > On 02 6, 07, at 8:06 PM, fooler wrote: > ... > whats the use of superior scheduler if your network stack sucks? of > course they work for it to improve it but still the overall > performance, stability, reliability and security what an OS can > give to you what you are looking for your server... > .... they're ALL equally important to get good performance, to get stability, reliability of an operating system. i.e. a sucky scheduler will reduce overall system performance like a sucky memory manager would... etc. those are some of the important elements that make up whats important in an operating system.
_________________________________________________ Philippine Linux Users' Group (PLUG) Mailing List [email protected] (#PLUG @ irc.free.net.ph) Read the Guidelines: http://linux.org.ph/lists Searchable Archives: http://archives.free.net.ph
