COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	           A PLAN FOR STANFORD'S NEXT GENERATION COMPUTING
C00009 ENDMK
C⊗;
           A PLAN FOR STANFORD'S NEXT GENERATION COMPUTING

	Stanford University has rather good computing facilities for
batch processing, and they are not yet seriously overloaded.  On the
other hand, the facilities for interactive computing, the use of
large data bases, support of small computers, and interaction with
experiments provided by the Computation Center are both inadequate
and excessively costly.  A large improvement in all these respects
is technically possible without a large independent development
effort.  (We could also have done much better with the present
generation, but this would have required development.)


	Consider the following plan
for next generation computing:

	1.  The  University Computation Center acquire an IBM 370/168
and a Digital Equipment Corporation KI10 computer system.  If there is
much delay in implementing the plan, a faster version of the KI10
will probably be available.  The KI10 is about the same speed as the
Computation Center's IBM 360/67, but it has much more advanced
interactive software than any IBM computer, and this advantage is
likely to continue because of the heavey concentration of general
purpose interactive computing on this equipment.

	2. Each of these systems  should  be  operated  in  the  most
advanced time-sharing mode provided by its manufacturer.  In the case
of the KI10 this is probably the Tenex system, and in the case of the
168  it  is probably the variant of CMS currently recommended by IBM.
The Computation Center should not modify these systems to  the  point
where  adopting  further manufacturer sponsored improvements requires
substantial work.

	This has not been a realistic option until recently especially
with the IBM software, but now that IBM has adopted paging on all 370's
there is a good possibility that IBM will converge on reasonable
interactive hardware.

	3. The two systems should interact by a direct connection and
through  sharing  files, but neither should be considered a satellite
of the other. Both systems admit this degree of  interaction  without
doing violence to the manufacturers software .

	4.  As  soon  as  it  becomes  technically  and  economically
feasible,  the  Computation  Center  should  operate  a  campus  wide
keyboard  and display based terminal system.  Current technology will
permit terminals that can be located in faculty and graduate  student
offices  and will permit undergraduate access through terminals rooms
in University public areas and in dormitories.  The system  currently
being  developed  for  the Computer Science Department can serve as a
prototype of such a system, and the  60 terminal 
system in use in the Artificial Intelligence Laboratory demonstrates
that suitable cost and performance goals can be achieved.

	5. The combined system should include random access bulk
storage (still disk in this generation of machine) at the lowest
feasible cost, because the direction of development of applications
is going in directions in which bulk storage is essential and not
just a convenience, and the cost of the storage is critical to these
applications.

	The advantages of this system are the following:

	1. Most of the programs in the world are  developed  for  and
available on the IBM 370 computer.  However, IBM has still not solved
the problem of adequate interactive computing at a reasonable cost.

	2. The best  and  most  economical  interactive  software  is
available  on the PDP-10 computer, and further such software is being
developed at Stanford and many other  places.   The  PDP-10  lead  in
interactive software is likely to continue for a number of years.

	3.  In  this system, the 370 can be used interactively either
with such software as IBM provides or indirectly via the PDP-10.  The
indirect  access  can  be either by having files edited on the PDP-10
put into the 370 batch stream or by  having  the  PDP-10  simulate  a
number of interactive terminals for the 370.

	4.  The  terminal  system  connected  to  both  machines  can
accomodate whatever relative demand turns out to exist.

	We regard this scheme as superior to one based on a dual  168
system optimized for the needs of the Linear Accelerator Center.