[email protected] (John Gilmore) writes: > Page selection and selective page rewriting operations are, or at > least should be, entirely separable operations. Moreover gather-write > channel programs are in my experience 1) as efficient as > connected-block ones and 2) not notably more difficult to write and > test.
re: http://www.garlic.com/~lynn/2014i.html#96 z/OS physical memory usage with multiple copies of same load module at different virtual addresses other page replacement trivia as undergraduate in the 60s, I did a lot rewritting cp67 for pathlength optimization, I also did dynamic adaptive resourcement management scheduling ... work on page replacement and paging optimization ... as well as changing the disk device driver to support ordered seek queuing (as opposed to FIFO) and page request chaining (on 2301 fixed-head paging drum, increased max thruput from 80/sec to nearly 300/sec) ... nearly all of this was picked up and shipped in standard cp67 ... even before I graduated and joined ibm. part of the page replacement involved work on "global" replacement algorithms ... at a time when there was a lot of academic literature about "local" replacement algorithms. at Dec81 ACM SIGOPS, Jim Gray asked me if I could help a co-worker with his Stanford PHD (Jim had left IBM research fall of 1980 to go to Tandem). His coworker's PHD involved "global" page replacement and Stanford was under intense pressure by the "local" page replacements forces (dating back to late 60s). Jim knew that I had significant amount of data for global/local page replacement comparisions, both implemented on cp67, showing global significantly outperformed local ... and he hoped that the actual data would provide Stanford with enough material to standup to the "local" page replacment forces. A problem arose that IBM research management refused to give me permission to respond to Jim's request (even in spite of the fact that most of the work had been done as undergraduate in the 60s, before joining IBM) ... which went on for nearly a year. Part of response that I was finally permitted to send, but not until nearly year later http://www.garlic.com/~lynn/2006w.html#email821019 I was being blamed for online computer conferencing on the internal network in the late 70s and early 80s and I hoped they thought they were somehow punishing me for online computer conferencing ... and not taking sides in the academic dustup at Stanford over local versis global page replacement. past posts mentioning page replacement http://www.garlic.com/~lynn/subtopic.html#clock past posts mentioning online computer conferencing http://www.garlic.com/~lynn/subnetwork.html#cmc -- 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
