Re: NetBSD vs FreeBSD

["Ian D. Leroux" <idleroux%fastmail.fm@localhost>]

My apologies for the hideous line-wrapping in the message I just sent. 
Fixed below.

On Tue, 02 Aug 2011 21:12 +0200, "Daniel Carrera" <dcarrera%gmail.com@localhost>
wrote:
> Hello,
> Still curious about the differences between NetBSD and FreeBSD, so
> Googling around I came across this post:
>
> http://mail-index.netbsd.org/port-i386/2009/02/09/msg001189.html
>
> The poster says that NetBSD is faster.

Such sweeping performance claims are nearly always silly.

> Does anyone here agree/disagree? I ask because I thought FreeBSD was
> supposed to be the speedy one. In general I don't care much about
> speed, but some times I run simulations on my computer. I have no
> idea if any BSD or Linux is particularly good for that.

Historically, the received wisdom was that FreeBSD was more heavily
and effectively optimised for i386 performance (especially server
performance) than NetBSD.  FreeBSD put a lot of effort into SMP in the
early-mid 2000s which initially did bad things for single-core
performance (the infamous 5.x series, now long behind them) but gave
them a head start on effective use of multi-core processors.  NetBSD
put a lot of work into this area more recently and has mostly caught
up.  A very impressive series of presentations came out along with the
NetBSD-5 release to publicise this fact, showing NetBSD outrunning
FreeBSD on a number of standard benchmarks; I'll bet the post you link
to was about them.  They laid to rest the idea that NetBSD was
generally slower, but I don't think they prove that NetBSD is
generally faster. Even if it was at the time the benchmarks were run,
both systems are constantly evolving.

Remember that many of the speed comparisons you see online are for
servers, or benchmarks that try to approximate server workloads.  The
requirements for number-crunching simulations are likely to be quite
different, making those benchmarks more than usually useless.  For
instance, in the limit that your simulations are CPU-bound the OS
should make no difference at all.  In reality, differences in memory
management and disk i/o will probably have an influence, but it's very
difficult to guess how much without a careful benchmark of the
workload you care about.

In practise, proper configuration of your system is likely to make a
much bigger difference to performance than choice of OS.  That, in
turn, implies that you should look for good documentation and a
community of people using the OS for purposes similar to your own, so
that you can learn how to get the best out of it.

Ian Leroux