Re: [SAGE] Servers per SysAdmin

["Mark C . Langston" <mark@bitshift.org>]

On Tue, Oct 09, 2001 at 09:22:33AM -0400, Trey Harris wrote:
> 
> You'll still need sysadmins, but the ratio will be dramatically lowered if
> the sysadmins are only putting out fires and handling hands-on work and
> not having to worry about ongoing builds, patches, upgrades, etc., as
> well.  Thinking of the sites I know of, with a strong architecture staff
> you can get the ratio down to 1:600 or more with a homogeneous
> environment, and 1:300 or so with a heterogeneous environment.  But this
> is definitely not the "turnkey" or "shrinkwrapped" solution that types who
> listen to Gartner tend to go for. :-)

Degree of homogeneity is a factor, but two items I see left out of
these discussions consistenty whenever they arise are: Job description
and interdependency (though I believe both have been touched on
obliquely in this thread).

Job description can be critical in determining the server:SA ratio.
I've seen many environments resembling those Trey and others have
described -- largely homogeneous, well-architected, etc. -- in which
many sensible people would say a ratio of 100:1 or higher is
appropriate.  And in some of these environments, it is.  In others,
it's woefully inflated, because the SA's role includes dealing with
users and management as well as the systems themselves in outage
situations.  In environments where the SA is "wind, watch, and wait"
-- i.e., the SA is tasked and left to her own devices until she can
report that the work is complete -- high server:SA ratios work well.
In environments where the SA must interact with (hundreds or thousands
of) angry or confused affected users while they're also expected to
correct the source of the anger and confusion, these high ratios don't
make as much sense.

In short, it's not just the technical environment that is a determiner
of the server:SA ratio, but the management and process in use as well.


System interdependency can be a killer as well, and to an extent this
hinges on what you define as a homogeneous group of systems.  If the
homogeneity extends beyond hardware and platform to functionality,
then you probably have a low degree of interdependency within that
group.  I.e., if there is a problem internal to a system, it will
likely not impact other systems of that group.  If there is a problem
external to the group of systems, the impact will typically be
all-or-none.

In environments where the SA is expected to oversee a number of
interdependent systems, the ratio should be kept low.  The reason for
this is somewhat obvious: Though it may seem that, in calm periods,
the SA could easily maintain a high number of these systems, when one
thing breaks it tends to affect other systems for which the SA is
responsible, compounding her problems.  This creates the "hours of
tedium punctuated by moments of panic" scenario we all know and
love...none of the systems are entirely down, but a non-fatal problem
with one creates a cascade of similar non-fatal problems with the SA's
other systems.

Of course, the worst-case scenario based on this is an SA expected to
manage interdependent, heterogenous systems in a high system:SA ratio
environment, where management and process dictate that the SA must
also perform helpdesk-like duties for her systems as well.

This seems to be more the norm than the exception.



-- 
Mark C. Langston
mark@bitshift.org
Systems & Network Admin
http://www.bitshift.org