In article <1990Dec13.011756.22888 at phri.nyu.edu> roy at alanine.phri.nyu.edu (Roy Smith) writes:
>> We're thinking about future directions of scientific computing here
[ deleted stuff about current facilities at PHRI]
> So, what I'm interested in, is how do things work in other places?
>Is the whole concept of a "central computing facility" outdated? If such a
>facility should exist, what form should it take? If not, should there still
>be some sort of central service department that supports individual personal
>machines, and provides a network to connect them to each other and the
>outside world? Provide shared peripherals like printers? Centralized
>backups? File servers? User consulting?
>>Roy Smith, Public Health Research Institute
>roy at alanine.phri.nyu.edu -OR- {att,cmcl2,rutgers,hombre}!phri!roy
There are lots of reasons why central facilities will never be outdated.
Almost from the very beginning of the personal computer era, there have
been a lot of things that are done _better_ by dedicated machines (eg. word
processing, handling of data from sequencing projects, use of
spreadsheets), but there are other things that are still best done in a
centralized fashion. These would include:
a) Database work - Databses are moving in the direction of weekly or even
daily updates. As the 'matrix of biological knowledge' becomes more
interconnected, it becomes virtually impossible, and undesireable, to
simply put everthing on the PC.
b) Maintanence and distribution of new programs, especially non-commercial
packages that may need some work to adapt to the local system.
c) Access to networks. Email and bulletin boards require a lot of disk
space, and some administrative work. It is also useful to have centralized
FTP directories and servers for distribution of programs and data.
d) Processing requiring more computing power than PC's can deliver. (eg.
intensive use of global database searches, phylogeny searches, multiple
sequence alignment.
Centralization
- eliminates duplication of efforts in obtaining and maintaining programs
and data
- decreases costs for software, where site licensces or licences for
distributed use are available
- simplifies distribution of updates: the data is all in one place
- fosters communication among lab groups that use the same data or programs
- decreases the need for personal disk space
- makes it cheaper, in the long run, to have sophisticated computing
capabilities
- makes it possible to have centralized support
Here at the Univ. of Manitoba, use of our central computing facilities is
provided at no cost for research purposes. This makes it easier to set up
software and databases for public use, and promotes continuity of services,
since each researcher doesn't have to worry about getting renewed funding for
computing in each grant proposal.
In general, what is really needed at each campus/institute/research center
is an individual who thinks that centralized computing for molecular
biology is essential, and who is willing to take on the administrative
tasks, as well as maintainence of software, data and documentation that are
necessary for such a system. Without such a person, the system will never
work well. When I was a postdoc at North Carolina State, we had two ATT
3B2 workstations, but no support money for those systems. A campus-wide
network was not supported by the administration. We also had the GCG
package running on a VAX, but there wasn't enough disk space to keep it and
the databases all on line. The University had bought a new VAX with
immense amounts of storage, but the administrative hurdles to use of that
machine prevented it from being used for molecular biology. In short, all
of the pieces were there, but there was no one with the motivation and
clout to make it into a useful system.
Because the Universityy of Manitoba has a strong commitment to centralized
computing (including a campus-wide network), it was almost trivial for me
to set up a molecular biology computing resource in the first few months
since my arrival here.
On our central SUN/Unix system, we now have several major databases and at
least 100 programs for sequence analysis available to all researchers on
campus. Since all of the programs we are using are public domain programs
(eg. Pearson programs, MBCRR programs, Phylip, FSAP, XYLEM etc.) the only
funds that were necessary was for the purchase of databases themselves. But
even that will soon be decreasing, as we move towards updating by FTP.
To reiterate, then, I think that two elements are necessary for a
centralized computing facility: an administration that promotes
centralization, and at least one individual with sufficient motivation and
clout to make it happen.
===============================================================================
Brian Fristensky | I'm not making this up:
Dept. of Plant Science | ... MY FRIEND FLICKA was pulled from Grades
University of Manitoba | 5 & 6 optional reading lists in Green Cove
Winnipeg, MB R3T 2N2 CANADA | Springs, (Fla.) after parents complained...
frist at ccu.umanitoba.ca | The book contains the word "bitch", in refer-
Office phone: 204-474-6085 | ence to a female dog, and the word "damn".
FAX: 204-275-5128 | Winnipeg Free Press, 30 October 1990
===============================================================================
