COMMENT ⊗   VALID 00007 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	.device XGP
C00004 00003	.s Goals
C00007 00004	.s Plan
C00010 00005	.ss Multiprocessor System Development
C00015 00006	.s Facilities
C00017 00007	.onecol s Budget
C00018 ENDMK
C⊗;
.device XGP;
.require "bask.pub[sub,sys]" source_file;
.sides←1;
.require "twocol.pub[sub,sys]" source_file;

.MACRO BC ⊂ BEGIN PREFACE 0; INDENT 1,4; CRBREAK nojust ⊃
.
.MACRO BS ⊂ BEGIN PREFACE 0; INDENT 1,4; nojust ⊃
.
.MACRO IB ⊂ turn on "%";
.AT """" ⊂ (IF THISFONT=1 THEN "%3" ELSE "%1"); ⊃
.AT "<" ⊂ "%2" ⊃;  AT ">" ⊂ "%1" ⊃;
. ⊃

.MACRO BIB  ⊂  CB(References);
.	BEGIN INDENT 0,3; NOJUST; IB;
.	AT "AIM-" ⊂ "Stanford AI Memo AαIαMα-" ⊃;
.	COUNT exref TO 200
.	AT "⊗" ⊂ IF LINES<3 THEN NEXT COLUMN; NEXT EXREF; ("["&EXREF&"]  ") ⊃
.	⊃
.SECNAME←"";
.portion some; place text;
.s Goals
The Stanford Artificial Intelligence Laboratory proposes to develop an
efficient operating system for the S-1 computer over a period of about
three years.  This proposal covers the first year's work.

The proposed work will have the following subgoals.

1.  Within one year, develop an interim operating system that services a single
S-1 processor and several display terminals, providing full timesharing
service.  Disk file access will be provided through a peripheral processor
and will not necessarily be fast.  This system should work
efficiently with compute-bound jobs, but not necessarily for I/O-bound
tasks.  Programs written for this system will probably have to be
extensively modified, at least in their input/output sections,
before they can be run on the multiprocessor system.

2. In parallel, design and begin development of an operating
system for multiprocessor S-1 computer configurations with their own
disk systems.
This system will provide efficient resource allocation for configurations
of one to about 32 processors and will include user interactive facilities
that are optimized for display terminals.
Teleprinter terminals will also be supported.
In addition to the operating system, a number of utility programs will
be developed, including text editors, file management programs,
compilers, and debuggers.

3. Based on work done in pursuit of the first two subgoals,
recommend specific equipment characteristics needed to support
efficient operation.
This particularly includes the manner in which secondary memories
and peripherals will connect to the system.

.s Plan
The two operating system projects will be undertaken in parallel by
separate groups.  There will be little formal coordination of these
efforts, but frequent informal communication among participants
will be encouraged.

.ss Interim System
The interim system is expected to use a modified version of
the Unix timesharing system running on a PDP-11 computer connected to a
single S-1 computer.
The PDP-11 will provide all interactive terminal functions
and, by suitable manipulation of a PDP-11/S-1 interface,
will load jobs in the S-1 and supervise their execution,
providing whatever input/output services are needed by the S-1.

It will be possible to run the Pascal and Fortran compilers
and the assember being developed by another group.
The interim system will be instrumented to permit timing of various
functions and debugging of equipment.

The general approach will emphasize getting a simple version
running as quickly as possible, then enhancing its
capabilities until it adequately supports initial testing requirements.
There will be little attempt to optimize performance
or to make it a useful production system.

From this point on, the personnel involved will provide only maintenance
support for the interim system and will join in the longer-range development
project.  It is anticipated that this task will be completed in less than
one year.

.ss Multiprocessor System Development
The advanced system will exploit the capabilities of multiprocessor S-1
configurations and will use the better features of existing timesharing
systems, such as Unix, Multics, TOPS-20, ITS, and the Stanford Monitor.
There will also be some innovation in interactive user services.

A key problem to be solved is efficient allocation and scheduling of
resources.
It should be possible to flexibly allocate processors either to a number of
tasks supporting independent users or to separate forks of a single task,
depending on priorities.

Interactive user services will be provided by one or more peripheral processors
and will feature a display editor as the top-level process.
This process (known in various systems as the "shell", "exec", or "monitor")
will permit the user to prepare text, initiate processing of multiple tasks,
and monitor and control execution.
The latter functions include dealing with error conditions from the various
tasks and subtasks under his control.

To aid in these functions, the user will be able both to partition his
display screen (to show information from multiple processes), to
switch one or more segments of the screen among processes, and
to direct keyboard commands to the various processes under his control.
The display control functions must provide the user with enough context so
that he does not have to spend a lot of time figuring out what he is looking
at.

The planning phase of this work (about three months) will be devoted the
following tasks:
.bs
(1) familiarization with the S-1 equipment characteristics;

(2) characterization of the principal kinds of computing tasks that are to be
performed with this system;

(3) general design of program services to be provided by the operating system,
including primary memory allocation and file system characteristics;

(4) general design of user services, including display control, command
languages, and character set standards;

(5) analysis of other resource allocation issues;

(6) formulation of criteria for selection of programming languages to be
used in major development tasks.
.end continue
Products will include an assortment of planning documents and specific
recommendations on equipment design issues, such as how
the disk memories should be connected.

The subsequent design phase (about six months) will focus on detailed
design of the functional elements of the system, selection of system programming
languages, and programming of developmental
tools (e.g. simple editors and debuggers), and possibly the modification
of a compiler to produce S-1 code.

In the following three month period, intensive programming on various
system elements would be underway.

.s Facilities
Much of the planning and preliminary programming work on both projects
will be performed on the existing computer facilities of the Stanford
Artificial Intelligence Laboratory.  Since this equipment has already
been purchased, mostly with U.S. government research funds, the only
costs involved in its use will be the support of part of a computer
technician and a share of other maintenance costs.

In support of the interim timesharing system development,
a multiprocessor PDP-11 system will be needed and is to be
made available on loan from LLL.  This system will consist
of two or more LSI-11s sharing XXXk words of memory and a
ddd disk.  It should be made available as soon as possible,
but not later than 1 April 1978.

The budget includes $UUUUU to cover licensing costs for acquisition
of the Unix timesharing system, which is to serve as the basis
for the interim operating system.

In addition, an S-1 processor with YYYk words of memory and
a PDP-11 interface is to be provided on loan from LLL by about 1 July 1978.
A second S-1 processor and some additional memory will be needed
later to permit development and debugging of the multiprocessor
operating system.
.onecol; s Budget
.once center
One year beginning 1 January 1978
.<< REG, JBR, BH, 2 studs, 1/2 sec, 1/2 tech.
.>>