Aqumix.1058
net.general
utzoo!decvax!ittvax!qumix!msc
Fri Apr 30 11:41:37 1982
Breakthrough
The following item appeared in the IEE News dated April 1st, 1982.
"BRITISH TEAM IN ECONOMICS BREAKTHROUGH"
"A DRAMATIC NEW BREAKTHROUGH IN THE EFFICIENT USE OF LARGE-SCALE
COMPUTING SYSTEMS HAS BEEN CLAIMED BY A GROUP OF BRITISH SYSTEMS
THEORISTS. EMBODIED IN A RADICALLY INNOVATIVE COMPUTER OPERATING
SYSTEM, THE NEW BREAKTHROUGH, WHICH IS TO BE FORMALLY ANNOUNCED
ON THE DAY OF PUBLICATION OF THIS ISSUE, PROMISES TO
REVOLUTIONISE THE WHOLE CONCEPT OF COMPUTER OPERATION BY APPLYING
TO THE PROBLEMS OF OVERALL SYSTEM CONTROL THE LATEST DEVOLOPMENTS
IN ECONOMICS THEORY.
To be known as the Total Optimisation of Resource Interaction
concept (TORI), the new approach is aimed at solving one of the
main problems facing those responsible for the efficient
operation of large-scale computing systems and networks --
that of ensuring that the computing resources available are
shared between the various application programs so as to
optimise the operation of the system as a whole.
Algorithms
Currently available systems all approach this problem of
efficient resource sharing by the use of complex
resource-allocation algorithms contained in a large and often
cumbersome piece of software known as the operating system.
Behind these resource-allocation algorithms there is
invariably a more or less complicated concept of pre-planned
priorities so as to maximise throughput, with each functional
unit in the system being loaded according to its capacity and
each application program being allocated system resources
according to its needs.
It is exactly this concept of pre-planned priorities that has
been so dramatically rejected by the TORI researchers. Such a
concept, they point out, leads to gross inefficiencies in
practice, not least because of the considerable overheads that
are introduced by the need for detailed administration of all
system resources by the central operating-system software.
Only the applications programs 'owned' by the system's users
actually contribute to its useful throughput, the TORI
theorists point out, so all overheads introduced by the
central (or 'publically' owned) operating software represent
nothing more than a waste of system resources.
`The trouble with conventional operating systems', a TORI
spokesman pointed out, `is that they spend too much time
worrying about how to allocate system throughput and not
enough on how to maximise it. What we need to do in future is
to spend less time in working out how to divide up the 'cake'
and more on how to make it bigger.'
The concept that the TORI team intends to use in order to
fulfill this aim was revealed recently at their research
centre in the Buckinghamshire town of Milton Friedman. The
idea is essentially to scrap the concept of resource-allocation
by pre-planned priorities and to replace it by one of
allocating resources by direct transactions between
applications programs and the system's functional units.
Transaction Units
In practice, each applications program will be credited with a
number of transaction units, and each functional unit will be
given a target number of transaction units that represents how
much it needs to 'earn its keep'. From then on the whole
operation of the system will be based on direct 'buying and
selling' of system resources on what the TORI group calls a
'free-market' basis.
As an example, if under this system two applications programs
simultaneously require access to, say, a line printer, their
order of execution will be decided not by the 'public sector'
operating system but by direct bidding within the 'private
sector'. Thus the line printer will make itself available to
the applications program that is able and willing to 'pay' the
greatest number of units. As a result, the TORI theorists
point out, the most productive programs will automatically be
allocated the most system resources, optimising throughput for
the system as a whole.
Overheads
An important additional advantage of this approach, according
to its proponents, is that it allows the 'bureaucratic'
overheads of the operating system to be greatly reduced. In
fact, this 'public sector' software can be scrapped almost in
its entirety, with nearly all its remaining functions being
'privatised' by being incorporated into competing
private-sector utility programs that can 'hire themselves out'
to other private-sector programs.
`All this is much more than just theory', pointed out the TORI
spokesman, `in fact, the TORI principles have been under test
on a major national system for nearly 3 years now.' When
asked whether the predicted large-scale benefits had yet been
achieved in practice, the spokesman added that certain
external factors had as a matter of fact meant that the real
benefits had not been achieved so far, but there was now every
sign that what he called `the upturn' would occur in the very
near future.
Leaked
Privately, however, an informal group within the TORI team
have made known their doubts about whether the radical new
operating scheme is actually capable of achieving the
predicted results. Known as the Weak Economic Theory Subgroup
(WETS), this part of the TORI team has 'leaked' a number of
disturbing facts about the actual operation of the TORI
experiment.
According to these leaks,the TORI system has been beset with
recurrent problems of what is known as 'inflation', which
apparently involves the system's functional units constantly
increasing their transactional thresholds or 'prices', while
applications programs constantly demand greater transactional
budgets or 'wages'. In addition, there is said to be a
growing problem of the system's functional units being unable
to find applications programs capable of keeping them fully
occupied, leading to them being designated as Disconnected
Operations List Entries, or being 'put on the DOLE'. In
terms of memory units alone, this phenomenon of resource
'unemployment' has apparently reached the unprecedented
proportion of over 3MBytes.
As well as these doubts raised by the WETS researchers, the
TORI team also faces competition from a number of other
groups. Their main competitor is in fact still very much
committed to the concept of resource allocation by pre-planned
priorities, but is said to be having continued difficulties in
making the component parts of its system work together
convincingly.
New Approach
These problems within TORI's main competitor have recently led
to a new approach being promoted by a recently formed joint-
venture development between an existing small team and a new
'spin-off' operation headed by four ex-members of the main
anit-TORI group. This new joint venture development has
recently been attracting a great deal of interest as a result
of its claims to be able to give everyone the best of both
worlds, but it has been noted that, everytime its spokesmen
are asked to explain their system's resource-allocation
policies in detail, no real explanation is forthcoming.
-----------------------------------------------------------------
gopher://quux.org/ conversion by John Goerzen <
[email protected]>
of
http://communication.ucsd.edu/A-News/
This Usenet Oldnews Archive
article may be copied and distributed freely, provided:
1. There is no money collected for the text(s) of the articles.
2. The following notice remains appended to each copy:
The Usenet Oldnews Archive: Compilation Copyright (C) 1981, 1996
Bruce Jones, Henry Spencer, David Wiseman.