Reading this thread is somewhat amusing.  I think people have been
reading the best practices docs for so long they've stopped thinking
about why the docs say what they do.

  Adding a server to put the DC on separate hardware from Exchange
does not increase availability.  In fact, it *decreases* the
availability of Exchange.  With the DC on a separate server, if you
lose the DC, you also loose Exchange.  So now you've got twice as many
failure points as before.

  I've seen this called the "airplane rule": A twin-engine airplane
has twice as many engine problems as a single-engine plane.  Sometimes
you really want that second engine (like on a plane), but don't loose
sight of the forest for the trees.

  Having a separate DC increases availability if you have *multiple*
separate DCs.  That's a real gain, since if you loose either one,
everything keeps running happily on the other.  (Modulo FSMO roles and
GC.)

  Distributing services is always nice.  Ideally, we'd put DNS, DHCP,
WINS, Exchange, DC, file, print, and databases each on their own
server, and each application would have its own server, and we'd have
cold spares for everything, too.  But putting them together doesn't
cause instability in the space-time continuum; it just means more
things go down with that one box.  In a very small company, it's not
uncommon to find everything tied together *anyway*.  See the DC
example above.

  Of course, software conflicts and the like do exist.  In my
experience, Microsoft PSS is a bit schizophrenic about the whole
DC+Exchange thing.  They say you shouldn't do it, but then when you
ask, "What about SBS?", they say, "Well, okay, yah, there's really
nothing wrong with it, we just don't recommend it.".  They seem to
want it both ways.

  That said, I've run single-server environments for years, and I've
never really have much of an issue.  But I'm careful about my IT
management practices, and make nightly backups.

  Performance analysts will cringe at this, but a typical
single-server environment is handling maybe 20 to 80 users.  With
today's hardware, you can be in the mud in terms of theoretical
efficiency and still handle a workload that small.

  Another thing to consider: For most very small companies I've dealt
with, the plan for large-scale disasters (e.g., building burns down)
is called "chapter 7 bankruptcy liquidation".  Things of that scale
are simply more of a hit than they can absorb.  So there's no point in
spending money on that scenario, at that stage.  Just make sure you
re-evaluate that decision as the company grows!

-- Ben

~ Upgrade to Next Generation Antispam/Antivirus with Ninja!    ~
~ <http://www.sunbelt-software.com/SunbeltMessagingNinja.cfm>  ~

Reply via email to