An interesting snippet from Computer Shopper in the UK -

"LINUX BEATS WINDOWS KERNEL

Results from a Microsoft research program show that the Linux and FreeBSD operating systems are more efficient than Windows XP. The Singularity project is Microsofts attempt to develop a "robust and dependable" operating system from the ground up.
In illustrating how efficient this new operating system is, Microsoft highlighted how much faster open-source operating systems are compared with its existing commercial product. The researchers were no doubt pleased that the new singularity kernel was more than twice as fast as the Linux 2.6.11 kernel when creating and starting new processes. But Windows XP's performance was over seven times slower than Linux and five times slower than FreeBSD."

-------Computer Shopper Feb2006------


Interesting that the attempt to build a "robust and dependable" OS can be read as a tacit admition that the current range of products is anything but.

That MS's own research proves what the Linux community has been saying all along smacks of poetic justice.

I'm willing to bet you won't find that little tidbit on their "get the facts" FUD site.

great news and timely !! so ... what should we do when we are given this benchmarking facts ?? ... bright future ahead for linux desktops and servers ...

talking about benchmarking , heres another reference ... its a benchmarking done by someone few years ago on linux 2.4 / 2.6 , OpenBSD 3.4 , FreeBSD 5.1 and NetBSD 1.6.1 , and the last update for this page seems to be at around Oct 23 2003 , this guy also included his benchmarking tools and his webserver softwares for downloading and with sources for building ... , seems that he was using an io library from D. J. Bernstein , probably these tools and softwares are still working , so maybe people down here can try it themselves ...


. Benchmarking BSD and Linux

. gatling server - his webserver and benchmarking softwares

note :: probably your *nix distros also got the binary/source/package of gatling 0.8 or more ...

________________________________

unrelated but anyway :: heres is one of the site page of D. J. Bernstein , start from his main site , probably students down here can find something usefull out of it ...



enjoy !!

.

Thats only one specific task the new windows kernel is faster than linux at.

I'm sure when the new version of windows is in beta stage there will be a survey that produces the results - "Well Windows Vista was X times slower than Linux in such a respect, but the new version is going to be x times faster than Linux". Its all about trying to make Windows look better

actually hard to say ...

cause things like benchmarking is kind of "unreliable" in all aspects ... what i mean(in fact this is what i suspect) is whether you do benchmarking on server applications performance or just user softwares application performance ... but it seems that user softwares applications are getting bigger and bigger ... anyway with the current power of pc , users hardly care or feel much about the merits of the underlying os unless they are doing things like heavy multimedia processing etc etc ... infact benchmarking is really cryptology for me ^_^ , so i could be wrong with that ... maybe someone who are experienced in benchmarking will sheds some light on it ...


.

Thats true, we all know that software can be specifically written to do well in a particluar benchmark, but being a trade off like all these thing it will suffer in another area. So I wouldn't read too much into the twice as fast claim, but for MS to state that Linux is 7 times faster than XP is amazing. Its not like they would pick benchmarks that make Linux look good is it?

More important than any of that I suppose is that MS are now developing a stable and reliable OS and their using Linux as their benchmark! Thats got to tell you something

Benchmarking is a great source of sound-bites for press releases, but truthfully speaking, usually they are a very biased source of information. Biased by their essential nature, which is to set up an artificial test-environment and test very specific variables.

If you were contemplating an actual purchase decision that could be thrown entirely to Linux, entirely to (ick...) Windows, or somewhere in-between, then you would have to much more systematically measure a whole lot of things; including many subjective things.

Like... that whatever-it-was that just went blazing past you looked a lot more like a penguin than a window ...

Here's the benchmarking of the firewalls of the free Unixes: PF (all BSD's), IPFilter (BSD, Solaris, HP-UX, etc.) and Iptables. It appears to be unbiased as it was done by an OpenBSD guy and the graphics show Iptables is really good:

http://insomnia.benzedrine.cx/pf-paper.html


By the way, Windows firewalls are a joke.

Has anyone ever benchmarked the comfort factor between city benches and bus station benches?

I don't know we have no idea what their set up was actually. For all we know it could have been a bbarebone linux system vs. a Fully Loaded Windows Xp system. We have no clue what they were evn using as a benchmark just what they were benchmarking for. So IMO not a lot of info really to justifiy going around being like "Linux is faster then Windows; MS proved it accordiing to this tiny exerpt in my magazine...."

I understand the argument the sometimes the bench mark is setup to favour one system or the other, and if it were a Linux house that was making the claim you could dismiss it. But is was a Microsoft project that said it. Are you seriously implying that Microsoft deliberately set up a test IN FAVOUR of Linux....I don't think so.

damn your good, thanks for the heads up, i would hate to wait 2.5 months before the arcticle in question was actually published before i could start laughing at it ...

and wth is up with this benchmark? how fast to start new processes?!?! i have a fast gentoo computer that can start processes at a good pace, then there is a linspire garbage computer that takes (literally) at least 15 seconds to start up its web browser. Every application on linspire takes at least 3X longer to load then on my gentoo computer, and guess what? my family like linspire, they dont even notice the slow loading times (i know i hate slow loading tho) No one gives a rat behind about startup times for processes, people can wait a few extra seconds (or 30 extra seconds ... i know i cant tho).

Even if a process starts up 2X faster then what i have now on gentoo, so what? i barely have time to put the mouse in position to move the new window to a better place on screen, before the window pops up. If the window pops up the same second i issue a command, does it matter? i know it wont do anything for any desktop user. does a half a second to 10 seconds really matter for servers either? do they spawn 1000000 processes, or do they take a shorter, faster route to deal with 1000000 requests?

Well, SciYro, having used older machines, I can say that a responsive desktop always feels more usable than a non-responsive one. The seconds do make a difference. Maybe the milliseconds don't, but they all add up.

wow, wasn't expecting that stream of burning vitriol.

The post was pointing out that Microsoft themselves are developing a new kernel and are using Linux as their benchmark because they admit that their current Windows product is seven times slower than Linux is in kernel performance. I'd say that show Ms as the hypocritical corporate fraud it is considering the time and money they have spent in trying to prove how rubbish linux is.

I expect MS are glad to have nutters like you scaring potential Linux users off of boards like this with your foaming at the mouth rants. Calm down fella, your going to have a heart attack.

btw Most magazines publish way ahead of schedule to allow for distribution times. Subscribers like me get them by mail which means they are often very early. Most things in life have rational explanations, you just have to calm down and think about it.

i am calm, and no foaming mouth rant intended. Just wanted to point out that process startup times are normally a bad benchmark, as most people only start processes in the GUI every few minutes, and in the commandline as much as one process a second. either way, most people, from my experiences, seem to prefer faster process completion, and better process smoothness and reaction time in the GUI, and rarely care about startup (which, btw, just because a kernel starts a process doesn't mean the user is aware of it, the kernel could be 10000000X faster at process startup, and user will not notice a thing, as most time is spent in the process loading and running itself before it can display itself to the user (at least of GUI process)