Very Computer

Can anyone help with to get a rough estimate of the number of systems they
manage.  We are trying to get some info on this.  Currently work in a group of
9 admins supporting around 400 Solaris, Irix, HP-UX, AIX, Linux and FreeBSD
systems.  Plus performing backups on around 500 PC's.  We don't have to support
Microsoft apps.  That is another group.  Are there any web sites out there that
have this kind of info.  Any help would be very much appreciated. Thank you
very much.

Jim Saunders

Unix System Administrator III

I would say that this setup is rather stressful.

Doing the math, this comes out to be around 45 servers per Admin. on a full set of
9. If they go on three persons per shift, then this would be 133+ per admin.

I am assuming that they are on shift. The most - less stressful and can give better
attention to details - an Admin. takes would be about 50 servers max.

Ed

:Can anyone help with to get a rough estimate of the number of systems they
:manage.  We are trying to get some info on this.  Currently work in a group of
:9 admins supporting around 400 Solaris, Irix, HP-UX, AIX, Linux and FreeBSD
:systems.  Plus performing backups on around 500 PC's.  We don't have to support
:Microsoft apps.  That is another group.

We have 2 3/4 people supporting 18 active low-end SGI workstations,
1 high-end SGI workstation, 3 SGI multiprocessor servers [20 CPUs
between them], 2 Solaris, 1 HP-UX; also 1 Netware server, and ~150 PCs.
Backups are handled nearly automatically for all the workstations and
servers {IMHO, the tape autochanger was one of the best investments
we've made}. We do have to support Microsoft applications. We're
also completely responsible for the network, and for security.
These last two make our situation difficult to compare directly against
yours.

Our Systems Administration and User Support group breaks down as:
- 1 PC specialist who also does some network work;
- 1/2 to 3/4 time backup Unix admin and backup PC, helps with networking
and security; the rest of their time is spend on research programming
- 1 manager / primary Unix admin / primary network / primary security /
postmaster; also sometimes does research programming, sets budgets,
evaluates technology, deals with policy issues, and everything else
computer-related for the organization. That's me, and it's more than a
full-time job.

Unfortunately, I do not have good estimates for the proportion of
my time spent "just plain administering" the Unix systems. My
network and security duties take up most of my time right now; recently
it was my research programming that took up most of my time. Depending
on how you define your terms, convincing arguments could be made for
ratios anywhere from 10:1 to 75:1.

Quote:>Doing the math, this comes out to be around 45 servers per Admin. on a full set of
>9. If they go on three persons per shift, then this would be 133+ per admin.
>I am assuming that they are on shift. The most - less stressful and can give better
>attention to details - an Admin. takes would be about 50 servers max.

It all depends on how they are managed.  If you use tools like Ignite/UX
to install them all identically and follow procedures to keep them all as
identical as possible, you can manage many more than if each is installed
independantly off the media (or worse yet, use whatever the factory
shipped them installed with) and each is treated differently with regard
to configuration files, installed software versions, patches, etc.  Now
obviously you can't say "let's standardize on Oracle 8i for all our Oracle
systems" because you have different requirements for layered applications,
but there's no reason at all you can't have maintain the same patches, the
same versions of stuff like Glance, ServiceGuard, Ignite and so on.

--
Doug Siebert

If at first you don't succeed, destroy all evidence that you tried.

: >I am assuming that they are on shift. The most - less stressful and can give better
: >attention to details - an Admin. takes would be about 50 servers max.

: It all depends on how they are managed.  

Yes, this is the key. Managing N identically equipped and configured systems
is not all that much more time consuming than managing a single system - at
least for the value of N in tens or low hundreds. The other important factor
is the type of workload handled by the systems: batch-processing systems
tend to require less work than actively used interactive workstations.

Regards,

/Serge.P

---
Home page: http://www.cobalt.chem.ucalgary.ca/ps/

At the moment, I'm running (solo):

(4) Solaris 2.6 (Ultra2)
(1) Solaris 2.6 (Ultra 10)
(1) Solaris 2.6 (Enterprise 3000)
(1) Solaris 7 (x86)
(10) Linux (x86)

We're ramping up to a lot more.  The configurations are dramatically different;
some are SMTP boxes only, some are name servers, some are log collectors and
statistics boxes, some are staging servers; in the future, some will be media
servers, others FTP servers, we'll have Oracle, MySQL, and Informix all running...
[eep]

I've always thrived on challenge.  =)

We have one UNIX admin (me), an NT admin, two LAN admins and a desktop support
analyst, for an office of approximately 100, and a whole slew of disparate servers.

Chris

--
Chris Harshman

+1-310-840-8967

I doubt you could get a sensible answer to that question from any one
person (including me - but I'm going to contribute my 2 cents anyway).

All of the information you desire depends on what your customers are
doing with the systems, what the systems are (servers or workstations),
and how competent your administration is.  Personally I prefer the rule
of thumb that if your administrators are basically not busy then you
probably have too many, and if they are overcome in times of emergency
then they aren't competent enough (even a gang of overschooled but
unintelligent admins will get overcome in an emergency because they
can't think on their feet - but again this is my own opinion).

:I doubt you could get a sensible answer to that question from any one
:person (including me - but I'm going to contribute my 2 cents anyway).
:
:All of the information you desire depends on what your customers are
:doing with the systems, what the systems are (servers or workstations),
:and how competent your administration is.  Personally I prefer the rule
:of thumb that if your administrators are basically not busy then you
:probably have too many, and if they are overcome in times of emergency
:then they aren't competent enough (even a gang of overschooled but
:unintelligent admins will get overcome in an emergency because they
:can't think on their feet - but again this is my own opinion).

I've seen a LOT of competent administrators overcome in the face of
incompetent environments.  It takes more than just the ability to
adroitly think on one's feet to deal with some kinds of real-world
catastrophic risks.  Appropriation of resources and priorities to
allow a sysadmin to do their job plays an important role.

--

 Royal Oak, Michigan | (has my PGP & Geek Code info) | Phone: +1 248-848-4481

True, very true.

heir feet - but again this is my own opinion).

--

All of these comments are mine
and do not reflect The Boeing Company's
policy or opinion.

It depends more upon the way the systems are setup than it does on
the competancy of an admin or how the systems are used.  

Quote:> Personally I prefer the rule of thumb that if your administrators
> are basically not busy then you probably have too many, and if they
> are overcome in times of emergency then they aren't competent enough
> (even a gang of overschooled but unintelligent admins will get
> overcome in an emergency because they can't think on their feet -
> but again this is my own opinion).

Someone else mentioned this, but I feel it does need repeating.  This
rule of thumb doesn't take into account the environment admins are
working in.  And as such it's not fair on the system admins at all.

In one environment that I worked in, I was one of two administrators,
and my primary responsibility was to design and build the security
architecture of a home grown e-commerce application.  The primary
Unix Administrator had to take care of more than ~50 developer systems,
Ultra5s and Ultra10s, and a farm of servers, 3 E3500s, an E4500, half
a dozen E450s, dozens of Ultra 2s, 7 Ultra 10s, a couple AIX systems
and 4 Linux servers.  That's over 100 systems supported by a single
admin who was also tasked with managing backups for all but the
mainframe systems.  The backups took up 75% of his time, my security
project 90% of mine, and together we spent only a portion of the rest
of our time managing the rest of the systems, and most of it working
on other projects.  

Another environment I worked in had two very competant admins who
had to work heavy amount of overtime managing their projects and the
20 or so systems they had.  I worked on an ERP setup there, and didn't
help with daily administration tasks.  They were no less competant
Unix administrators than either of the admins in the other
environment.  They just had to deal with much more legacy systems with
much less support to make large changes.

Starting from scratch with planning, a single admin can handle a
large number of systems.  It's rare that anyone gets that kind of
opportunity.

Best Regards,

Andrew Garman

Quote:> Can anyone help with to get a rough estimate of the number of systems they
> manage.  We are trying to get some info on this.  Currently work in a group of
> 9 admins supporting around 400 Solaris, Irix, HP-UX, AIX, Linux and FreeBSD
> systems.  Plus performing backups on around 500 PC's.  We don't have to support
> Microsoft apps.  That is another group.  Are there any web sites out there that
> have this kind of info.  Any help would be very much appreciated. Thank you
> very much.

I manage (solo) a cluster of 11 SGIs and 2 xterms, as well as a few
other SGIs.  In the end, if you do things intelligently, there should
be no difference between maintaining 1+N machines and 1+(2N) machines
(where N is the number of clients, and the 1 is their server, etc).
Every change I make is done to all clients together.  Of course, this
only works for each type of architecture.  In your case, you've got
6 architectures, so assuming a server and clients for each that gets
up to the equivalent of 12 machines (if managed intelligently), in my
book.  By comparison, I'd say I effectively manage 4 machines (a
server and clients for 6.5.6, a 6.2 standalone, and give help on a few
5.3 standalone's).  But I'm doing this in my spare time for free as a
hobby, and you're (presumably) getting paid to work full time....

Damian Menscher
--


--==## Physics Dept, 1110 W Green, Urbana IL 61801 Fax:(217)333-9819 ##==--

I once read a some sort of mathematical setup for this. It depended of
your experience, number of servers, number of users and so on. The
probem is that I lost the url to this document. If I'm not wrong, this
was from the sysadmin FAQ.

-Preben-

Quote:> Date: 19 Sep 2000 11:27:49 GMT

> Can anyone help with to get a rough estimate of the number of systems they
> manage.  We are trying to get some info on this.  Currently work in a group
> ....Solaris, Irix, HP-UX, AIX, Linux and FreeBSD   systems.

To get a good answer you need to be quite clear what
"manage" means and what the service level agreements are.

Also things depend on your customers and the tools they have
to get into and out of trouble (root/administrator access)
and training.

Your question prompted me to look at a corner of things
here.  In my area I see about 3500 systems supported by
about 15 part time folks and two manager types.  Perhaps
this works because of who we hire ;-).

But, as I look about the company I see wildly different
ratios.  The more Managers, VP's and marketing folks the
more support is needed ;-).

In this group it works because of the bounds we set and the
resources we have made available.  Folks tend to have enough
privledges, disk space, memory and CPU to do their job.  
They are also responsible for their files on their own work
station.

So part of any 'management' process needs to start with
attention to resource allocation.  Expensive shared
resources require management and planning!

We do have some very big machines (very much in demand) in
various groups.  These systems are sometimes managed
directly by a large staff of three or four full time people
often 7x24.  For the most part our groups large systems are
'open' to the group with /etc/motd and /etc/issue being the
key management tool.

Folks that do not play fair get 'managed'.

Culture is a key management tool.  Do help desk calls read
like "My disk, it ran out of space .vs. I filled up my
disk". The difference is a window into the personal
responsibility and involvement of your users.

As we know the culture of the internet has not improved of
late.  Well placed and managed firewalls can set bounds
within which folks can work and be responsible.

Above I said that "Folks that do not play fair get
'managed'".  This is in contrast to managing the asset.
Often there are multiple handles to a management problem.  
By working with the people you get to understand the need.  
By focusing on the asset you might miss the need.  Both are
important!