Douglas W. St.Clair <[EMAIL PROTECTED]> wrote:
> Peter L. Phillips wrote regarding his concern with an email send from
> within MS that may relate to the NT LINUX test in which LINUX did very
> badly. He wrote in part:
>
> >[The msg may have originated in] ITG is the organization within Microsoft
> >responsible for maintaining the main servers, and the Microsoft-only
> >customized versions of released (and
> >unreleased) Microsoft products.
>
> It would not be surprising if MS used LINUX servers internally and this guy
> may be running one or more of them and had nothing to do with the test.
There are quite a few people there running linux on spare boxes, but usually they're
just that, older, smaller
machines. However it is unlikely that someone would get a budget request approved for
a $20,000 computer with which
to goof around with linux . It's far more likely to be approved as part of a server
oriented product group's
testing or benchmarking budget.
In other forums, people have pointed out the similarities between this message and the
report in question. The
timeframe, hardware, problems reported, linux distribution, and 2.2.x kernel are all
in sync. I felt that people
should take note of where this message came from.
<Wild Speculation>
Consider this scenario:
A whizbang server oriented product group (WinNT, SQL Server, IIS, or what not)
receives a truckload of new Quad PIII
Xeons with 3GB for their performance lab. The old PII Xeon sits idle in the corner.
During a lull in testing
generic Server product 2000, someone decides to informally pit NT vs. linux. He
already has some baseline #'s for
that NT on that box, so he installs Redhat Linux, gets the latest kernel sources,
builds an SMP kernel, and starts
running a few simple tests using a small hand full of the client machines in the lab.
Wow, guess what? Because he
used poorly supported hardware, with a kernel that contains bugs that pertain to that
specific hardware config, with
improper Apache and Samba configurations, NT beat the pants off of linux. He uses
dejanews to request assistance,
but never provides additional information to the people who respond because he forgets
to go back on dejanews and
look for responses. He forgets to look on dejanews because he fixed the most glaring
problem with his linux config
and believes that he's been successful in tuning linux (and Samba & Apache.) He brags
to his coworkers about the
defeat of linux, and at the next team meeting the topic of his informal benchmark
comes up. Somewhere along the
line some thinks, "Wow, we've gotta tell the world, lets get someone to do an
independent benchmark." This thought
goes up the food chain to the VP for approval. After getting approval, the group's
manager arranges to foot the
bill for the Mindcraft benchmark. Mindcraft is loaned the server for the duration of
the benchmarking operation,
and the Microsoftie who did the informal benchmark is used as their linux tuning
contact.
</Wild Speculation>
> Peter have you tried to write to the source of this message and get more
> information?
No, I'll leave that to the journalists. In my own mind I'm sufficiently confident
that the person who posted the
message was in some manner involved in the benchmark.
The question is, would Microsoft's direct technical involvement in a benchmark that
they financed change anyone's
perception of the benchmark? I suspect not, so I will write nothing more on this
subject.
Peter Phillips
-
Linux SMP list: FIRST see FAQ at http://www.irisa.fr/prive/mentre/smp-faq/
To Unsubscribe: send "unsubscribe linux-smp" to [EMAIL PROTECTED]
