On Wed, 2003-04-30 at 20:33, dantrams wrote:

> Ok, I'll give on this point (that an idle Windows machine has more
> services
> running than a stripped Linux machine), but if we're talking about a
> fraction of 1% of the CPU resources difference between the two OS's
> at idle,
> how does that translate into a significant performance gain for Linux?

For one you also have a memory issue. What's windows smallest idle foot
print? You can run Linux in a server env without X using between 64-32
MB of memory or less and see decent performance. It all depends on your
use for the machine.

I run 756Mb in my Linux DB server and it get's used. FB,IB like most
Linux services can operate with good performance with limited resources.
The more the better as always.

Aside from that do get more specific networking is part of the core of
Linux. You can go so far as to compile your network interface driver
into the kernel itself. Along with what ever else specific to your
hardware, removing anything that is not specific. Loaded code is fast,
compiled in code can be faster.

To go further from my experience and others Linux makes a better network
device. Specifically I have use Linux as a router, firewall, and
switch/hub all with excellent performance. Some with performance that
rivaled or beat equipment from Cisco and others.

So I would go so far as to say in my experience certain networking tasks
can be done faster on Linux than Windows. Most Unixes as that level
withing reason will be faster. Although I have had a report from a
colleague that is an engineer in Cisco's fiber division. They swapped
out a Sun machine for a Linux one, and the network related task sped up
almost 10 fold.

Point, I have also seen better network TCP/IP I/O on Linux than Windows.

Of the wall point. If you were really nuts, you could have your FB
server use a RAM disk as the hard drive. Sure you can do that with
windows as well, but it's not as easy to get windows itself to operate
on a ram drive. I have working with Linux booting off a floppy, zip or
cdrom and running everything else after than entirely out of memory.
Talk about speed.

> Most or all of the IB engineers that were at Borland when Bill Karwin
> and Paul Beach and the others resigned are still with Interbase
> today (Charlie Caro, Siriam, etc).

Interesting, as I have received help from Paul years ago when I was
first learning about Linux and InterBase. He did not mention his
thoughts on the matter to me. I would be curious what his thoughts are
today. Not sure if he will read this thread or not.

> >
> > I would ask them. If you do not take their word then take no ones.
>
> I did -- I gave you their response which you apparently disregard.

I only disagreed due to my real life experience. I am sure you would do
the same if you were in my shoes. When I switched from a Win2k server to
a Linux one that was not a dedicated DB server I saw a performance gain.
When I built and dialed in a server specifically for Firebird, I saw
impressive performance gains. Did I benchmark them, no. Was the
difference obvious and great, yes. Were clients happy, yes.

I respect others opinions, especially those with more experience and
knowledge. But I saw what I did, thus I know what I do. Maybe not in all
cases, but definitely in mine.

> They did not show me any benchmarks either. They have an agenda (to
> sell
> Interbase regardless of the platform), as do you (to put down MS).
> In
> both cases, I take the claims with a grain of salt.

The only reason to Like Windows or Solaris users over Linux, is at times
in the past the ones using Windows or Solaris usually had bigger
budgets.

If you were a Sun or Windows server admin, who's company is paying out
the rear for licenses, and possible machinery. Just to find out a
product screams on hardware of half the price, and a free OS.

Many companies did not believe, and the present state of things is
reflecting the change.

So from a sales or salesman perspective you want to appear neutral to
all to move your products. If you are bias to Sun, Windows and Linux
users may be turned off, and visa versa.

Neutral is the safest route. If you are in the US you should remember
the whole Enron/Anderson issue. The builders should not always be the
ones to provide the benchmarks. Sometimes uninvolved third party
agencies can be more reliable at benchmarking.

Does this make sense or off the wall?

> I run a mix of Windows and Linux in my office.

Great run the benchmarks for yourself. Or forget specifics play around
and see for yourself. Try and do it from an open minded prospective.

> My personal
> experience is
> mostly with Windows.

Mine was to, 6+ years Windows and Dos, 3+ years Linux. Would I go back,
never. I do not have the justification. I have one Windows machine for
certain development purposes like Flash. I dual boot my laptop as well,
mostly do have access to some windows dll so I can run Wine in Linux.

> I'm open to switching platforms based on a
> measurable
> gain; for my purposes,

Then run some tests, as I would assume you do before making platform
based decisions for your company. Maybe you have easier access to
similar hardware than I do at the moment.

> cost is not as much of an issue so the fact
> that Linux
> is free and runs in 32 MB is not significant.

Cost I look at as money able to be allocated to additional hardware,
personal or anything else my business needs. Business could always save
money and or re-allocate funds. Using a free OS is one way.

With a free OS you also get something called Freedom. Something the U.S.
was founded on.

Freedom to have access to the code. Freedom to dial in and tweak things
for your needs. From to remove, and modify, and what ever else freedom
means to you. For a business it also means control. Control to not have
to depend on another company or person if you have the abilities, since
you have the access.

Now to address the 32Mb point. I am not suggesting you use a machine
with that. However if a machine performs well, or in acceptable range
with limited resources. Think of what could be done with larger or
unlimited resources.

A windows machine with 32Mb running FB as a server would perform like
crap. A Linux one would not. Beyond that is just math. Faster hardware
is just multiples of your performance difference on limited hardware.
Also if software can perform just the same on limited hardware, you have
no need to spend crazy amounts of money on hardware. What you do spend
you will get your moneys worth.

So you can call it savings or in other words efficiency.

--
Sincerely,
William L. Thomson Jr.
Support Group
Obsidian-Studios, Inc.
3548 Jamestown Ln.
Jacksonville, FL 32223
Phone/Fax 904.260.2445
http://www.obsidian-studios.com