[seul-edu] Floppies on Workstations

[Michael A Hamblin <michaelh@utdallas.edu>]

Hi folks,

In one of the labs here we have approximatly 30 PCs running Linux, however
the hardware is rather old and doesn't perform anywhere close to the level
we were hoping for (probably the reason the PCs were retired from the
Windows lab). What we are looking at doing in the next week or so is
setting up a Dell server with Linux and have all the Linux boxes pull
their sessions from that machine, at first just experimentally to see if
it will handle the load well.

One of the issues that was brought up was workstation floppy drives. I
know some people have dealt with this problem here, where you have
workstations and want to share out their floppy drives so that the server
can read them. We were looking at just piecing together some scripts to
handle all that but it seems hackish to me. Plus we're having a problem
where unmounting the shared floppy will return a device busy even though
nothing is accessing the floppy, and the only way we've figured out how to
fix this problem is to restart nfs (probably not optimal). If you've dealt
with this some feedback would be appreciated.

This configuration we're looking into was inspired by the success we've
had with the SunRay's (or in some cases lack thereof). Before I
transferred from the Engineering School to my current job in the IR Unix
department early this month, we set up a lab of 40 SunRay's. At the time I
was very happy with the configuration, as the SunRays were much more
responsive talking to the E3000 than the Sun Ultra 1's they were
replacing.

But as in so many things there were problems. Some of the tools refused to
run properly on the SunRay hardware (mostly X issues with Cadence and
Planet) though normal programs ran fine. This wouldn't have been an issue
but the timetable we were on was hyper-accelerated and didn't leave
adequate time for testing. The other problem was that due to a lack of
proper planning the E3000 was inadequate for the load being placed on it,
additionally no restrictions were in place to limit users processes.
Considering the heavy use of the lab it's not surprising that something
would blow up.

Fortunatly for me, the week before everything went haywire I had already
transferred to my new position in another department and I was merrily
enjoying a vacation in Illinois :) But the point is that the SunRay's are
really nice hardware when they are set up properly. The workaround
solution, and we'll see how this goes, is to purchase 2 420r's from Sun to
serve the SunRay's, where one of the machines will be a failover unit in
case the first machine is overloaded or otherwise crashes. Hopefully this
will aleviate some of the problems they have experienced. Additionally
they are looking in purchasing or leasing a small number of Ultra 60's or
80's for the expressed purpose of running heavier software packages.

We'll see how this works as it seems a bit like a bandage approach right
now. Some of the problem is probably the expectations that are placed on
computing equiptment in higher education. Information Resources here at
UTD is as expected at a University woefully underbudget (though not as
badly as we used to be) and is struggling (with a large degree of
success) to implement a campus-wide switched 100Mb network amoung many
other exciting projects. Individual schools and departments are
responsible for equiptment purchases and upgrades. Some schools, like the
"wealthy" Engineering and Computer Science school have been able to
provide decent running machines for faculty and staff (there was a major
push this spring to replace P166 or slower machines on desktops with new
PIII systems), however other departments (perhaps due to inadequate
funding or different funding priorities) plod along with really old
and inadequate hardware.

Another interesting thing I experienced in the Engineering school was that
the expectation of the support group was that the labs were to be
maintained through grant monies individual research groups
received. Unfortunatly not all the faculty think this way and have been
using other resources (such as the lab of Ultra 1's) in ways that the
support group was not aware of. One of the problems that arose from the
replacement of the Ultra's with SunRay's was that a number of professors
were using the lab machines for doing network simulations. After some
verbal parrying the solution was to take some of the Ultras that were
bound for Faculty desktops and press them into use as a network simulation
cluster, affectionatly known as the ballpark (before we had pitcher and
catcher, and the new machines were named for the other positions).

What I found interesting about this was though there were quite a few labs
that were receiving grant money and upgrading their computing facilities
as needed, there were still labs plodding along on five-year old failing 
hardware, and some faculty had the impression that it was the support
group who was supposed to go through and upgrade the equiptment. A very
interesting problem :) I think, at least in the Engineering school, that
there has finally been enough communication on this that this isn't a
problem anymore, but at the same time there remains the problem of
extremly old hardware waiting to fail.

Another interesting problem I ran into which was fairly well isolated to a
particular group in the Engineering school was a group of resourceful
people who would scavenge the surplus department for very old computing
hardware (P75s) and do a lot of work on the systems and make them
functional. Though I applaud their hard working spirit and their
resourcefulness, the problem is that eventually the support group will
have to support these configurations should the resourceful people leave
and then you have to wonder if a P75 is really worth the hours of work
that these people put into them. Certainly there are situations where you
would want to salvage old hardware however given the funding level of the
group this seemed to be a waste.

Anyhow, these were some of the lessons and experiences from working in the
Engineering school, so I am passing them along in the hope that someone
will get an insight into what can happen before embarking on a project :)

--
Michael Hamblin           http://www.utdallas.edu/~michaelh/
michaelh@utdallas.edu     http://www.ductape.net/
UTD Linux User Group      Technical and Communications Services
                          "Why would I want any more?" -Stavesacre