COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	TOWARDS MORE INTERACTIVE FILE BASED COMPUTING FOR STANFORD
C00010 ENDMK
C⊗;
TOWARDS MORE INTERACTIVE FILE BASED COMPUTING FOR STANFORD


	The subcommittee on research and  instructional computing for
the Stanford  campus has just begun its  activities.  We had expected
to spend  several months  determining our  opinion on  which way  the
research  and   instructional  interests  of  Stanford   faculty  and
students  can  best be  advanced  in the  next  computer procurement.
However,   it  seems  that  our hand  is  being  forced by  events  -
specifically by  what appears  to be a  decision to allocate  part of
the cost  of  an IBM  370/158  procured  at the  instigation  of  the
Medical Center  to general campus  computing.  Part  of a 158  may be
just what  the campus needs,  but  since the configuration was chosen
with no input from us, we are very doubtful. Besides our  worry about
the results  of the  decision, we  are also  quite unhappy  about the
procedure.

	The following ideas should be given serious consideration:

	1.     The  main  direction  in  which  we  should  look  for
improvement in  Stanford's next  computer procurement  is in  greater
interactiveness.

	2.   This  conclusion is  tentative.   It is  not based  on a
market survey  and  it probably  shouldn't  be, because  the  present
users of computers  cannot be expected  to see the  possibilities for
improvement fully.   The development of the  steam shovel was not the
result of demand  by a committee  of ditch diggers.   However, it  is
possible to  survey those parts of  Stanford such as IMSS  and the AI
lab  that have more  interactive systems and to  survey those schools
that are ahead of Stanford in interactive computing.

	3. We should expect  to develop an ever increasing  number of
computer terminals until almost  every faculty and student office has
one. The Computer  Science Department  has already decided  to go  in
that direction.

	4. Besides  inexpensive terminals,   it is important  to have
inexpensive computer files.   The IBM 3330 provides a basis for this.
The rent for 1,000,000 characters from IBM is about $5.00 per month.

	5. It is important  to have an  IBM computer available in  an
IBM system, because much of  the national academic community will use
this  machine.  IBM  is making progress in  interactive software, but
still is far from having a system that will  permit economical use by
a large community.

	6.   The leading community  in interactive computing  is that
based on the Digital Equipment  PDP-10 computer, especially the  ARPA
supported projects.

	7. It  makes  sense for  Stanford to  have both  IBM 370  and
PDP-10  capacity,   and  the ability  to take  advantage  of software
developed for both machines.

	8. We propose for consideration a configuration  including an
IBM  370 (possibly  the 158  already  here) and  a Digital  Equipment
KI-10 or possibly KL-10 if the latter is available in time.

	9.   Each computer will operate with the software supplied by
its  manufacturer   and  will  remain   sufficiently  loyal  to   the
manufacturer to  accept improvements in the operating  system as they
are made available.

	10. The two computers will share an IBM 3330 disc file.  Both
operating systems allow this  sharing.  Each computer will  also look
like a terminal to the other.

	11. Besides  this, the terminal system will  permit a user to
connect himself to either machine according to his needs.

	12. The  initial  emphasis  will  be on  the  PDP-10s  better
interactive  facilities  and  the   370's  better  batch  facilities.
Neither computer will be considered a satellite of the other.

	13.   The cost of the proposed KI-10 system would be $700,000
plus or minus $100,000 not  counting the terminal system or  the 3330
disk system.   A full stand-alone system  with these facilities would
come to  $1,400,000 including  a TIP  (terminal interface  processor)
capable of connection to the ARPA network.

	14. The operating  costs depend strongly on  whether Stanford
finds  it  possible  to  operate  a  PDP-10  without operators  in  a
Computation Center environment.   Other  PDP-10 systems are  operated
this way but not in a  computation center environment.  Because a lot
of money depends on it, this needs careful investigation.

	15.  We  estimate that  a  KI-10 will  be able  to  give good
service to 100  simultaneously logged  in users and  a KL-10 to  200.
However, this  depends a good  deal on what  the users are  trying to
do.

	It should again be emphasized  that the subcommittee has  not
had time  to  reach definite  conclusions, but  the subcommittee  has
reached the consensus  that this direction should not be precluded by
decisions taken at the administrative level.