Re: [tlug] VMWare GSX Server

[Patrick Niessen <tlug.niessen@example.com>]

On Thu, 13 Jan 2005 14:46:16 +0900, Stephen J. Turnbull
<stephen@example.com> wrote:
> >>>>> "Patrick" == Patrick Niessen <tlug.niessen@example.com> writes:
> 
>     Patrick> On Wed, 12 Jan 2005 11:54:00 +0900, Stephen J. Turnbull
>     Patrick> <stephen@example.com> wrote:
> 
>     >> >>>>> "Patrick" == Patrick Niessen <tlug.niessen@example.com>
>     >> writes:
> 
>     Patrick> No doubt that it is possible to run GSX on Gentoo, but
>     Patrick> its a matter of liability and safety.
> 
>     >> OH MY GOD!!!!
> 
>     Patrick> Stephen I think you have a theoretical point there.
> 
> Yes, I am a theorist.  The point I am making is 100% practical.  You
> have unlimited liability for the reliability of your systems; your
> vendor's is limited.
> 
> It's possible that the point doesn't apply to you; if you look at what
> I wrote, you'll see that I already accounted for that.  What I
> objected to was the general implication that avoiding open source and
> using commercial software instead is (clearly) a matter of liability
> and safety.  For your ass in the short run, true, you won't get fired
> for buying SuSE or Red Hat.  For your company's in the long run, it's
> not clear at all.
> 
> Are you aware that Japan has the highest rate of business closure in
> the OECD, higher even than the notoriously Darwinian U. S. of A.?  The
> reason is that in Japan it is small organizations that buffer larger
> ones as much as possible.  And that is despite the much higher social
> and economic penalties paid for bankruptcy here.  (There's also the
> fact that many small businesses in Japan are hobby businesses, which
> survive only by decreasing the owner's net worth by 10%-20% a year.
> Hard to compete with someone who considers that acceptable....)
> 
>     Patrick> Yes it is true that one can work so hard and make it work
>     Patrick> in the end, but would it not be nicer if it justs works?
> 
> I hope your business planning is not conducted on the basis of
> pleasant fantasies like that!  Would you put it in those terms to an
> investor (whether a stockholder or your parent company)?

This is exactly the point of standards.  If both sides know to what
well defined specification they have to work, and they have testing
procedures in place that ensure a compliance with the standard, then
chances are you can plug one into the other and it works.  If you
insist on deviating from the standard to reach a potential higher
level as you propose, then chances are that there will be problems
(which can probably be fixed through your hard work). The latter is
more expensive from a management point of view, and introduces a
significant amount of uncertainty into your critical path.  Its bit
like Russian Cosmonauts improvisation skills vs meticulously planned
western space missions.

It seems to me that you think that if the result is excellent, then it
should be called good quality.  Alas, this is a dangerous
misperception, a more accurate measure of quality would be to evaluate
of how much compliance with predefined procedures there was during the
implementation, as quality is only achieved if a certain performance
level can be constantly maintained, and is repeatable.

The other question is that if you are such a great sysadmin to hack
everything together, what happens if you die or change company. 
Unless you have a grudge you'd want things to keep running.  And you
cannot expect everyone to have the same high level of linux skills. 
(Maybe he started with suse and would have to learn all that gentoo
stuff again).

> The "theoretical" point is that the sysadmins for whom everything
> "just works" (and have reliability stats that have been known to make
> young girls squeal[1]) are the ones who ask the questions I proposed.

Besides being very generalising here, I think you move away from a
theoretical point towards a theological one.  But you better save your
powder for a better opportunity.

> True, any commercial Linux has _most_ of the advantages of Gentoo
> ... but it doesn't have the advantage of your greatest expertise.
> Isn't _that_ what is going to count most at crunch time?

Fortunately I am so clever that I can use even two different distros. 
Bit like driving cars: In Europe I can drive right side with manual,
over here left side with automatic.  Its not rocket science.
 
>     Patrick> I think most professional organisations work to
>     Patrick> standards.
> 
> Up to the point where they want to be successful organizations.  Right
> there they switch to _setting_ standards (at least, trying to).  Of
> course the converse logic also applies: the organization has to do
> both to succeed in the long run, whether it starts as a "professional"
> organization or an "innovative" one.  Maybe you're in the "follower"
> department and can get away with only the former.  I can't judge that,
> except to note that you said "small" business (or "small" branch of a
> large business, maybe).

I refer you to my earlier point in regards to standards. This is not
about strategic software and business advantage.  I just consolidate a
few servers onto one.   Might just as well buy loads of commodity type
cheap servers and cluster and provide failover instead of using VMWare
- is one approach better than the other?  Anyway, my boss does not
judge my performance (and pays me) for changing all infrastructure to
great linux systems, but for more tangible things like, "don't need to
replace 3 old machines costing 400,000 yen, so saved more than 800,000
yen", or "was able to test disaster recovery scenarios without
purchasing additional hardware".

Anyway, by all means please contribute your inevitable last word, and
then please create your own thread about standards, work ethos, and
innovation.  My intention was to hear from people who are using the
same product on linux about their experiences, and to pick up some
hints about the configuration .. alas your contribution, while highly
appreciated, has performed rather poorly in that respect ;-).
-- 
Patrick Niessen