Re: athena

[Adel Hanna <hanna@graphics.lcs.mit.edu>]

I have three comments:

1- I realize the convenience of having athena lockers at the lab instead
   of remote logins to athena workstations, and I think I will go ahead
   with testing it on Linux. (icon.lcs.mit.edu will be the test machine.)
   However, almost all of my users are using athena either for email or
   course works. My previous experience with maintaining athena in the lab
   (with sgi) regular  crashes on daily basis and the only student that
   really used it was Matt Antone reported to me that the performance was
   extremely slow. at any rate, to keep track of the athena lockers usage
   on icon.lcs can you give an idea of  what are going to use it for?

2- I'd like to have sun hardware in the lab, and supporting one machine
   is  like supporting many. You configure one and copy your configuration
   to all machines. So I would say I need to put %10 of my time to
   maintain a new system like sun in the lab.

3- I'm in favor of let students maintain special project machines
   themselves; however, I'm really not in favor of letting students
   maintain shared lab machines.


On Sun, 2 Feb 2003, Seth Teller wrote:

> peter, i agree that athena would be a good thing to have in the
> lab.  the students maintained an athena SGI some years ago (i
> paid the athena fees that existed then), and got decent use from
> it.
>
> regarding the issues you raise (space, adel's time, and h/w
> budget).  space and money are cheap.  time is not (patrick, you
> can misquote me on that).  time, including adel's time, is the
> limiting factor here.  what you propose would increase the
> heterogeneity of the lab in two ways:  s/w (athena) and h/w
> (sun platform), and therefore increase his maintenance load
> far more than, say, adding another O2.
>
> one way to ameliorate this is for students to take the lead
> on the acquisition and install.
>
> adel, can you comment?  how painful would it be to support
> linux & sun athena in the lab?  would you be comfortable if
> the students were to do it themselves?
>
> so, here are a couple of suggestions:
>
>   1) for now, let's do a trial run with one linux and one SGI
> machine (not sun).  yes, SGI is degenerating, but there's still
> much use in that configuration, and setting it up should be
> easy for the students with little overhead for adel.  (adel,
> can you help them find space in the lab for two new heads, or
> one new head with an appropriate switch?)
>
>   2) get me some plausible numbers for a) a suitable fast
> sparc machine and b) person-hours to set up and maintain it
> as athena linux.  i'll discuss with adel where to go next.
>
> seth.
>
>
> To: bmg@graphics.lcs.mit.edu
> Subject: athena
> Date: Sat, 01 Feb 2003 21:15:31 -0500
> From: Luka <luka@MIT.EDU>
>
> there's been some interest expressed in having an athena workstation or
> two in the graphics lab.  i looked into this and here's the latest data.
>
> there is a zero-cost athena option available where you support the
> workstations yourself rather than having I/S do it (as used to be
> the required case a few years ago).  i sent adel links on how this
> works which he should be reviewing.
>
> on specific machine types:
>
> - sgi's are being desupported by athena, so you can expect the locker
> software and athena package support for them to start decaying rapidly.
> (the locker builds for sgi are already rather impressively crufty.)
> also, the sgi installer is a bit finicky, and will not install very
> politely unless it recognizes the exact disk hardware and such that is
> on the machine as very-close to the official athena config.  so sgi
> athena might be a bad idea.
>
> - linux is ok.  the installer is fairly robust and will pretty much
> "just work" on any modern pc hardware.  much locker sw works on linux
> these days but there are still some gaps, on various big packages.
> notably autocad does not run on linux.  notably inventor does run.
>
> - sun is probably the preferred platform.  pretty much anything that
> ran on sgi now runs on either sun or linux, with a tendency toward
> sun first for the big packages.  notably autocad runs on sun.  i do
> not know about inventor/etc.  note that the weight of solaris/athena
> is pretty heavy these days and while it will work on something as old
> as a sparc5 that is kinda painful.. at least an early ultrasparc is
> recommended for a comfortable install.
>
> so i think getting one sun and one linux athena machine in the lab
> would be a nice configuration at face value.  contingencies are:
> 1. space to put them somewhere
> 2. an ok from adel on maintainance and any I/S issues
> 3. budget for hardware
>
> (1) is a serious consideration i leave open on the table.
>
> (2) is waiting for feedback from adel.
>
> (3) is probably a non-issue for linux since pc's are utterly cheap.
>     for sun, i'd say if we just want to give it an initial try then
>     use a sparc5-ish machine just because they are virtually free
>     (heck i even have one at home), but if we know we want to have
>     this as a useful machine rather than a questionable space usage
>     then spring for a fully current ultrasparc model (they are
>     reasonably priced as serious workstation market goes).
>
> that's about it.  comments from others?
>
> -pl
>