Re: [TCLUG:5285] NTS4 vs. Linux study from Mindcraft.

[pani@frontiernet.net]

Scott Dier wrote:
> 
> Hello, If you havent seen on Slashdot allreaday --
> 
> Mindcraft has released a report saying that NT is faster for web and file
> serving.  Go read it.  Check out some of the graphs.  Like the apache one,
> where it just drops off in the middle of the graph.  Why?
I guess all that we need to do is prove them wrong eh ?  Which
would mean duplicate the tests and see how they measure up.

But then it would take the fun out of all the "discussions" that
this is going to spawn. :-).

I hope somebody from the "linux" turf with the know-how carries
out the same test in the immediate future and show us all "how 
wrong" mindcraft is. 

my 2c,
sandipan

> 
> Because Mindcraft is a bunch of inept sysadmins.  So when your boss goes
> NT is better than Linux, you can tell them about these facts.
> 
> Mindcraft turned off "widelinks" in samba.   From the original author of
> Samba he said that with it turned off it takes 6 more function calls to
> perform operations.  That sounds like it would probally run faster
> (Someone should get around and prove this.)
> 
> Secondly, the apache stuff can be explained by one thing : File
> Descriptors.  You can change them to higher numbers if you want, but
> Mindcraft left them at the standard number.  So while the computer *was
> not* hitting its max CPU, it was waiting for a file descriptor.  Mindcraft
> could have reconfigured the kernel and had /much/ better performance and
> see that Apache, while it might not be faster -- scales much better than
> IIS does.  I saw an article from a /long/ time ago on a zdnet site that
> showed apache scaling from 100 to a ton more (1000+ i think) as a flat
> line for latency of getting pages.  Thats cool :)
> 
> Really, I'm throwing this forth as a discussion object and as something so
> when the FUD hits the non-techies and they ask you about it you have
> something concrete that you know is right.
>