22 Jun
2016
22 Jun
'16
10:31 p.m.
On 06/22/2016 06:18 PM, Paul Koning wrote:
That's where SCOPE differed. A program could request a specific amount
of ECS, just like CM field lenght. At least that was the situation in
say SCOPE 3.1.6. ECS wasn't terribly useful in the early days.
Of course, under Zodiac (and TOOS), we kept whole program chains in ECS
shared among up to 4 systems. A chain could run in any machine and have
any of its modules accessed from any other machine or ECS directly. We
were very ECS hungry at a time when CDC wondered if anyone would be
willing to buy the stuff. We may have had the first 4MW ECS setup sold
to a customer.
None of our PPUs accessed ECS directly, however. Our database system
didn't use files either--completely random access across, IIRC, up to
144 844s online. Neither SCOPE or KRONOS was up to that job.
--Chuck
On Wed, 22 Jun 2016, Chuck Guzis wrote: ...
> Its illegitimate relative, KRONOS, made extensive use of ECS for
> support of the PLATO system.
Not quite. KRONOS treated ECS (rather clumsily) as a kind of disk.
PLATO just bypassed all that and managed ECS directly, as memory the
way it was originally designed to be used. After startup, PLATO
would own all the ECS and do transfers directly, without any OS
involvement. It would also use ECS for inter-job communication, and
for communication with PPU programs. For example, PLATO disk I/O
uses both request queues and data buffers in ECS, which the PPU
program accesses. Ditto for terminal I/O.