What is faster: freebsd or Linux?

Steve O'Hara-Smith steve at sohara.org
Thu May 21 19:24:26 UTC 2020


On Thu, 21 May 2020 20:23:58 +0300
nikitastepanov113 at yandex.ru wrote:

> What is faster: freebsd or Linux?

	Oh dear - time to drag this one (from 2006) out again.

HH> >> Which is better tea or coffee?
HH> >Beer.
HH> Scotch.

	THC.

	But seriously the question is meaningless unless you add some
specifics.

	Fictitious but reasonable demonstration of why:

	Two individuals A and B are arguing over the relative merits of
their favorite systems OSA and OSB. Both are happy with what they have, but
they agree to try the other and soon find that they were both right OSA is
better for A and OSB is better for B. But they want to know which is *best*
even though anyone with sense might stop at this point.

	Careful analysis reveals that OSA happens by default to be
configured in a way that suits the hardware and workload that A has, but
this configuration does not suit B's rather different setup and load as
well os OSB does. Experts are called in to tune both OSA and OSB properly
for each site. They test again. Lo and behold OSA is a clear winner, both
settle on it. The argument is over ... or is it ?

	B gets some new hardware and calls in the same expert to tune up OSA
for it. The performance is not what was hoped, just for fun he tries OSB -
it flies better than was expected for OSA. A gets involved again, they
realise that the two systems they originally tried had a major component in
common (motherboard say) and the new box has a different one.

	They spend a fortune and buy every type of processor, motherboard,
disc controller, disc drive, etc, etc. They set up a massive test with a
fancy benchmark suite and test OSA and OSB across all this hardware
variation and make a big graph (it only takes a decade or two but hey they
*want* to know). They find that OSB wins on most setups and OSA wins on
quite a lot, OSB is delcared the overall best. Settled ... or is it ?

	They publish their results in a million dollar report (got to pay
for that hardware somehow), and as soon as it is published everyone in the
industry criticises the benchmark as being totally unrealistic compared to
real world use. Other point out that almost all the hardware they tested is
now obsolete and the new stuff is very different.

-----------------------------------------------------------

	Yes, you really could go through all that and *not* answer the
question. I have seen smaller scale insanity performed quite often, usually
it sells magazines.

	However if you have a real job to be done and real hardware to do it
in, then just try both and use the one you wind up liking best - or pick one
and use it unless it barfs or irritates in which case try something else.

-- 
Steve O'Hara-Smith <steve at sohara.org>


More information about the freebsd-questions mailing list