From: "Joe Barr" 

On Fri, 09 May 2003 20:20:17 -0400, Geo. wrote:

> "Joe Barr"  wrote in message
> news:pan.2003.05.09.23.40.18.961501{at}austin.rr.com...
>
>> http://www.linuxworld.com/2003/0509.barr.html
>>
>> Microsoft claims Windows 2003 Server is twice as fast as Linux, at least
>> when it's used for file serving. I spoke to Jeremy Allison, head of the
>> Samba team, who provided a few insights into the test configurations
>> that don't leap out at the reader because they are hidden away in
>> appendixes to the benchmark document. Allison feels this, in itself, is
>> substantially responsible for the outcome.
>
> from the article:
>
>>"If you look at the curves in the benchmarks, what is really really
>>obvious
> is that Samba, or the kernel, isn't running out of steam," Allison said.
> "What's running out of steam is the disk subsystem." <
>
> I agree with him on the RAID0 setup it could easily be disk, I don't know
> anyone using RAID0 for a fileserver (isn't your data important?) but:
>
>>RAID 0 wasn't the only poor choice for Linux in the tests. "They used
>>ext3,
> which is one of the slowest filesystems on Linux," <
>
> I can't agree on this point, ext2 isn't suitable since it's so easy to
> wipe out with a simple power failure. In a fileserver you have to be able
> to count
> on the file system coming back up after a hard poweroff. Fileservers are
> where everyone stores their data, the file system is critical. ext3 is the
> only choice.
>
> I also don't agree with his other ramblings about unfair tweaks to the
> system as he pointed out above, the bottleneck was disk so any other
> tweaks are pointless, even with RAID0 the bottleneck is still disk.
>
> This leaves 2 questions in my mind, first how would they have compared if
> neither machine was running RAID0 but instead were running RAID1 and then
> running RAID5 (much more common configurations for fileservers) and second
> but perhaps meaningless to this test is why is RH's RAID0 so sucky?
> RAID0's whole purpose is for speed.
>
> Geo.

One other thing.  How about comparative benchmarks between W2K and W2K+3?
If they don't show the same kind of performance increase, that proves the
tests were rigged.

Do you know of any?




--

--- BBBS/NT v4.01 Flag-4