I just *had* to comment.

[dsr at tao.merseine.nu]

On Thu, Jun 03, 2004 at 03:35:33AM +0900, Derek Martin wrote:
> On Wed, Jun 02, 2004 at 05:07:09PM +0000, dsr at tao.merseine.nu wrote:
> > > As for why it's desirable, how many sysadmins do you think it takes to
> > > manage 1500 client machines and a couple dozen server machines
> > > 24/7/365?  The number might surprise you.
> > 
> > Exclusive of help-desk activities and specialists like DBAs, that could
> > be handled by a team of 6 or so. You need 2 people during the
> > day, 2 more to handle two additional shifts, and 2 to cover
> > vacations/sick days/conferences. Another one is probably
> > advisable, so I'll say 7.
> 
> HAHAHAHAHAHAHAHAH!  That's really really really really funny.
> 
> Just keeping windows up-to-date will probably be a full-time job for

What Windows boxes? We're discussing all-UNIX environments with
proper support infrastructure for system administration. We're
just disagreeing about what constitutes proper support
infrastructure.

> about a dozen people in such an environment.  Then on top of that, you
> should expect to have one admin for every 30-60 users, depending on
> the specifics of the environment.  Your comment suggests that you've
> never worked in a large environment before.

Possibly not. Are 15,000 servers considered small these days?

> I have re-implemented, or been involved in environments that
> re-implemented something very similar to what Derek describes on more
> than one occasion.  And I never lived at MIT.  Doing so, in as much as
> is possible at your site, makes everything a lot nicer.

That I find interesting. Did you use AFS? Was it a university?

> > The number of client machines is not important to anything
> > except hardware replacement and initial install. 
> 
> That's quite preposterous.  You don't work in support at all, do
> you...

Installing a new machine is a matter of:

- add entries in DNS and DHCP
- bringing the hardware in, plugging in cables, and setting the
  BIOS for PXE booting or CD booting.
- power-up; PXE boot and automatic OS and system installation,
  or alternatively CD boot and automatic OS and system
  installation.
- login; tell the machine what group it is part of; run the
  first configuration pull by hand to watch for errors.

Anything subsequent is either O(1) for universal changes, O(#of groups
affected), or related to hardware failure or handled by the help-desk.

-dsr-