Linux/BSD network performance comparison

Dear all:
This question has been in my mind for a while and I searched on the
internet trying to find the answer on my own. Unfortunately, I
couldn't find any document with 'data' to prove their opinion.

Can anyone show me some links about the network performance
comparison between BSD and Linux with some real data?

thanks in advance

tcs

"TCS" <(E-Mail Removed)> wrote in message
news:(E-Mail Removed) om...
> Dear all:
> This question has been in my mind for a while and I searched on the
> internet trying to find the answer on my own. Unfortunately, I
> couldn't find any document with 'data' to prove their opinion.
>
> Can anyone show me some links about the network performance
> comparison between BSD and Linux with some real data?

I have no imperical measurements, but I have run both OS's on the same
physical box. In cli mode, I'd rate BSD as perhaps 2% quicker on response
to external (web, ftp, etc.) requests. In the ease of installation
department, Linux (SuSE 8.1) was much easier to get the desired services
working. The hardware for this was an old K6/2-450Mhz, a D-Link 10/100 NIC,
512MB ram, and a 7200rpm 40GB drive.

If you are setting up a server with the expectation of heavy load, OpenBSD
would probably out-perform a Linux with 2.4.4 kernel by a small percentage.
I'm not convinced that it was worth the trouble to get everything working as
desired (particularly the NIC was problematic) with OpenBSD.

JW

On Tue, 28 Oct 2003 07:37:32 -0800, TCS wrote:

> Dear all:
> This question has been in my mind for a while and I searched on the
> internet trying to find the answer on my own. Unfortunately, I
> couldn't find any document with 'data' to prove their opinion.
>
> Can anyone show me some links about the network performance
> comparison between BSD and Linux with some real data?
>
Most data you'd get benchmarking Linux and the BSD's would normally arrive
with a full-swing of flame-festing comments (as neither side would want to
give in to the other). Hint: There was a recent benchmark for such
posted at slashdot. Guess what - it became a long thread full of hot
comments.

On my unqualified, unscientific, personal observation, having worked and
tuned up Linux and some BSD's (NetBSD, OpenBSD, FreeBSD), I would say the
following:
1. OpenBSD is a bit slower compared to the other three. I would
expect such slowdown as acceptable though, given that embedding crypto to
the OS would definitely slow down things.
2. The variation of network performance between the BSD's and Linux don't
really get too far from each other. Maybe it's because of my settings,
though.

At any rate, Linux as well as the BSD's are good operating systems, each
with their own merits. I just so happen to deploy a mixed setup of them
for the sake of "genetic diversity" and play with their mixed strengths.
--


Paolo Alexis Falcone
(E-Mail Removed)

Martha Stewart called it a Good Thing whenPaolo Alexis Falcone <(E-Mail Removed)>wrote:
> On Tue, 28 Oct 2003 07:37:32 -0800, TCS wrote:
>
>> Dear all:
>> This question has been in my mind for a while and I searched on the
>> internet trying to find the answer on my own. Unfortunately, I
>> couldn't find any document with 'data' to prove their opinion.
>>
>> Can anyone show me some links about the network performance
>> comparison between BSD and Linux with some real data?
>>
> Most data you'd get benchmarking Linux and the BSD's would normally arrive
> with a full-swing of flame-festing comments (as neither side would want to
> give in to the other). Hint: There was a recent benchmark for such
> posted at slashdot. Guess what - it became a long thread full of hot
> comments.
>
> On my unqualified, unscientific, personal observation, having worked and
> tuned up Linux and some BSD's (NetBSD, OpenBSD, FreeBSD), I would say the
> following:
> 1. OpenBSD is a bit slower compared to the other three. I would
> expect such slowdown as acceptable though, given that embedding crypto to
> the OS would definitely slow down things.

The issue with OpenBSD isn't so much crypto as it is that they use
aggressive strategies vis-a-vis memory management to fight against
buffer overflow attacks.

The way to think about it is to look at tools like Electric Fence and
Checker; OpenBSD winds up integrating this sort of stuff into their
LIBC, in order that conscious memory leaks are unlikely to be able to
smash their way anywhere useful.

EF/Checker are ludicrously expensive to run, so you don't run them in
"production;" the OpenBSD folks have built their memory tools to be
rather less expensive, but it doesn't come for free...

> 2. The variation of network performance between the BSD's and Linux
> don't really get too far from each other. Maybe it's because of my
> settings, though.

There's pretty regular competition back and forth; the improvements in
Linux 2.4 can be recreated in FreeBSD 4.6, and the improvements in 4.7
can lead to discovering further improvements for Linux 2.5.

In addition, the place where the BSD guys claim to do better is in
stability under _heavy_ loads; you may not necessarily be able to
replicate the heavy loads needed to see the differences.

> At any rate, Linux as well as the BSD's are good operating systems,
> each with their own merits. I just so happen to deploy a mixed setup
> of them for the sake of "genetic diversity" and play with their
> mixed strengths.

Genetic diversity is indeed a good thing. There are vulnerabilities
that we see ("Outhouse Express" and "Internet Exploder" being all too
typical) that result from the near monoculture of Windows. If the
world changed into a Linux monoculture, it would still be a
monoculture, vulnerable to all the things that that implies.
--
output = reverse("ac.notelrac.teneerf" "@" "454aa")
http://www3.sympatico.ca/cbbrowne/linuxxian.html
Jumping off a cliff doesn't kill you! It's only when you hit the
ground...