re:
http://www.garlic.com/~lynn/2012o.html#30 Regarding Time Sharing

oh, old presentation at fall atlantic share meeting in 1968
http://www.garlic.com/~lynn/94.html#18 

on MFT14 & CP67. CP67 had been installed in the univ last week of
Jan68. Univ. continued to run OS/360 (in 360/65 mode on 360/67)
production during the period. I was the support person for os/360
... but did get to play with cp67 some on the weekends.

most of the presentation was about cp67 pathlength rewrite during spring
& summer 1968 (before I really got in to totally rewritting virtual
memory and dynamic adaptive resource management). it does reference that
I had completely redone os/360 sysgen ... hand reordering much of the
stuff to achieve optimal disk operation (and while I was at it, changing
it so it could be done in the product jobstream)

standard sysgen out of the box took a over 30 seconds elapsed time (with
hasp, before hasp, it was well over minute) for typical univ. student
job workload (3step fortgclg, before watfor). This was mostly job
scheduler & allocation doing approx. 10seconds of loading linklib and
svclib members per job step. Careful sysgen reorder got it down to 12.9
seconds elapsed time (note quite three times faster).

the re-order accomplished two things 1) it group the highest used
linklib & svclib members together at the start of PDS dataset ...
minimize arm seek between PDS directory lookup and loading member and 2)
placed the highest used members at the front of the PDS directory. full
cylinder 2314 PDS directory search took 20 revolutions at 2000rev/min
.. well over half-second elapsed time during which time channel,
controller, and disk were locked out. Even half-cylinder pds directory
multi-track search would take 1/3rd second (per member load). Making
sure that the highest used members were on the first track or two of the
PDS directory ... could reduce PDS directory member lookup to 40-50
mills. 2314 reference from bitsaver
http://bitsavers.org/pdf/ibm/dasd/A26-3599-4_2314_Sep69.pdf

past posts on dynamic adaptive resource management
http://www.garlic.com/~lynn/subtopic.html#fairshare
past posts on virtual memory management
http://www.garlic.com/~lynn/subtopic.html#wsclock
past posts on getting to play disk engineer
http://www.garlic.com/~lynn/subtopic.html#disk
past posts on ckd dasd, multi-track search, fba (including
being told that even if I gave MVS fully integrated & tested
FBA support, i still needed an additional $26M new business
profit)
http://www.garlic.com/~lynn/submain.html#dasd


-- 
virtualization experience starting Jan1968, online at home since Mar1970

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN

Reply via email to