[ale] News from the front: The O/S Wars Continue

[hirsch at mathcs.emory.edu]

> Yes, this was a fair test. Competition is a Good Thing. This gives us a
> goal to shoot for; for instance, I'm sure the new IP stack will be
> significantly faster. It would be good to have a comparison once a year. I'm
> sure Microsoft and the Linux community would be happy to fund it.

Before you believe it too much, be sure to read
http://www.heise.de/ct/english/99/13/186-1/
which is from the German magazine c't.  They ran their own tests.  It turns out
that the main flaw in the Linux kernel is the way it handls multiple NICs.

When they ran similar test with only one network card, Linux was significantly
(though not enormously) better than NT.  But with 4 networks cards as in the
Mindcraft tests, NT was much better.

When they added in tests with dynamic content, NT really sucked compared to
Linux, though the test wasn't really fair since both systems used CGI (with
perl, I think).  I think there are better ways to do this for both systems, but
this was the only mechanism in common.

It looks like once we get a more reentrant networking code, Linux will win on
that test, again.  Basically, it appears that MS figured out the one place that
Linux code didn't work well, then designed a test to expose it.  According to
c't, few webservers are going to be heavily loaded on 4 NICs simultaneously.

Another intersting article is http://cs.alfred.edu/~lansdoct/mstest.html,
which demonstrates that even if you buy the Mindcraft numbers, Linux can still
fill 7 or so T1 lines for a much cheaper price than NT.  The c't article makes
similar observations.

--Michael

> 

> > On Mon, 28 Jun 1999, Keith R. Watson wrote:
> >
> > > "PC Week Labs' tests show what path Linux must take"
> > >
> > > http://www.zdnet.com/pcweek/stories/news/0,4153,1015266,00.html
> > >
> > > keith
> > > -------------
>