[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Sites Pilot

Well considering we just rolled this out (i.e. announced it etc.), we
aren't going to have any useful data to evaluate until Spring.   Additionally
with things frozen with CSCAR like they've been no one's been
spending any significant time doing any stats eval on these boxes.

      -- Andrew

On Fri, 21 Jan 2005, Gavin Eadie wrote:

> below is from September.  The pilot period is over so it's time to
> review it.  Who should work on this for me?  Mull it over as a group
> [add to agenda on 26th?] and let me have two or three names next week
> (by end-of-day Friday 29th).  Then I'll work with the selected team
> to close on this quickly ... Gavin
> ______________________
> From: Kitty Bridges <bridges@xxxxxxxxx>
> Date: Sun, 5 Sep 2004 08:56:39 -0400
> The UMCE statistics Solaris machines currently located in the Sites
> were decomissioned over the winter/spring.  Not only was the hardware
> outdated to the point of being unusable, but the OS couldn't be
> maintained securely and usage has dropped dramatically over the last
> year.  This is in part due to a shift away from solaris and in part
> because the hardware isn't particularly powerful.
> The plan is to eventually replace them with Intel hardware running
> linux.  This will provide better support for statistical
> applications.  We plan to use Sites provided hardware, and
> consolidate purchasing/support costs.  We plan to test this
> environment with the existing UMCE linux distribution.  Sites folks
> would provide a loadset based on this distribution without extensive
> effort.  It's an estimated 1-2 weeks to develop the initial loadset
> (mostly Sites time with assistances from GPCC folks) and 1-2 weeks to
> build overloads for statistical packages and test them.  An open
> source windowing application would be deployed with linux, without
> additional development.
> We hope to have 4 machines to a Fall pilot (2 Angell, 1 Ugli, 1
> CSCAR).  At the end of this Fall term pilot, we would evaluate:
>   = how well this environment (hardware, OS, application
>     software) works for statistical applications, and
>   = the level of support needed for this environment.
> After this evaluation, ITCS will decide whether or not to expand the
> pilot in its current state, or to make changes (in hardware or OS) to
> better support statistical applications users.  We will explicitly
> review the cost and benefit of using UMCE linux in the Sites setting.
> There are also additional collaborative opportunities that should be
> investigated relating to this, with CSCAR, Natural Resources and
> other departments with statistical labs and intensive usage.  This
> should be further explored in the Fall.