[HECnet] Another TOPS-10 DECnet problem
Thomas DeBellis
tommytimesharing at gmail.com
Thu Nov 11 14:48:13 PST 2021
Oh, that old cat fight? Meow!! I'm walking away from it; I don't know
how much email got spewed between MRC and BAH about it. I don't think
either side ever got the point that you are not comparing apples to apples.
Having looked at both schedulers, I don't immediately see that either
was more efficient than the other. There clearly was cross
fertilization in a number of areas.
Recall that Tops-20 has processes and that a job may have a large number
of processes. The number of jobs then is not going to be a valid
comparison. For example, let's take a look at Galaxy on Tops-10, which
occupies 10 job slots:
Job Who Line# What Size(P) State Run Time
1 [OPR] DET NEBULA 26+40 HB 0
3 [OPR] 0 QUEUE 9+38 ^C 1
7 [OPR] DET QUASAR 40+40 SL 1
9 [OPR] DET PULSAR 5+40 HB SW 0
10 [OPR] DET ORION 109+40 SL 0
11 [OPR] DET NML 15+18 HB 3
13 [OPR] DET CDRIVE 30+40 HB 0
14 [OPR] DET FAL-10 104+40 SL 1
They're all underneath a _single_ job on Tops-20 or built into the EXEÇ,
but producing the same load because it is the same code.
We did do some instrumenting and we found that the snazzy parsing
(COMND%) was not contributing that much to load. There was some
overhead simulating UUO's, which are obviously natively executing on
Tops-10. Nearly all editing was done with WYSIWYG video editing, which
surely must produce more load than TECO or SOS. Some work was put into
TEXTI% to mitigate the context switching.
MRC's position was that Tops-20 was doing more, but I'm not sure how
comfortable I am with that. Having used and programmed both, I think
it's more like 'doing differently'. I would say that it was rare to
find people who could easily move between the two and/or who weren't
highly opinionated.
It's a waste of time; you bought what did the job best for your
environment. It's kind of like apples and pineapples; they sound the
same but they're just not.
------------------------------------------------------------------------
On 11/11/21 5:20 PM, Robert Armstrong wrote:
>
> >You had a 20 that would handle 600 students in 1977/???/
>
> I think he said something about six 20s… I’m pretty sure there’s no
> way one CPU would have handled 600 timesharing users. We could get
> to around 120 on a single KL10E with TOPS-10 before it got unbearably
> slow. With TOPS-20 on the same hardware we could only get to 80 or
> so; TOPS20 was something of a pig.
>
> Bob
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.sonic.net/pipermail/hecnet-list/attachments/20211111/28756f1a/attachment-0001.htm>
More information about the Hecnet-list
mailing list