I was wondering if I should bother responding to 1kyle - decided to go to bed instead.
Sounds like someone who has read the books but neglected to read the (Linux) source.
"Swap" in Linux (despite the name) is nothing like swapping/page trimming in classic MVS.

foo_bar_foo covered most things I think.

I may be in error, but I think that what Linux calls "swapping" is what an MVSer would call "demand paging," and what an MVSer would call "swapping" or "anticipatory paging" does not exist.

if someone could enlighten us further
"swapped out" was the MVS term for paged out
and "paging" was certainly the MVS term for extended storage
but i don't think that extended storage was storage like a hard drive.
I think MVS is using the lingo to describe what for us is the difference between L1/L2 CPU cache and RAM
RAM in MVS lingo is "expented storage" ?
not sure.
I mean like a S/390 is HUGE almost like a super cluster or something to us desktop guys. So you guess this stuff was really big amounts.
back in the day i wrote some programms for MVS on punch cards and stood in line to feed em through.
did the think even have a hard drive. i guess it did ?
we certainly never tried to use it for anything.
possibly the hard drive or drives were its RAM.
MVS had segment table which was pointers to page tables
and page table with pointers to 4k pages.
it also had Translation Lookaside Buffer so it is more related to Linux paging than it seems on the surface.
address translation was done in hardware DAT like modern MMU
so "swapped out" was actually a bit like Linux swap but at a lower level.
Linux swap is related to demand paging in kind of the same way.
the differnce has to do with where the pages were going.
"swap" for MVS is like for us moving pages into physical ram from cpu cache.
? certainly not sure about this ?

Way off topic, and too big a call.
I might throw something at the wiki one day - start here for an (easy ... ) intro ...

from article above
i take that to mean real as in what was RAM at the time
128KB for early MVT
and absolute as some kind of either tape or disk.

Whoa - that link was for the latest and greatest follow-on to MVS (z/OS).

MVT was early '70s - everything has changed since then. MVT only knew of real storage - the concept of prefixing, and thus absolute addressing, came much later.
Wonder how the NUMA folks handle similar issues in Linux - must look one day.

i just mistyped
MVT was dysfunctional / actualy created partitions on the fly till it died of fragmentation
and evolved into MVS (in the 70s) not much later at all.
1973 introduction of virtual memory on IBM S/370 Models 158 and 168
i relize alot of what we are talking about is hardware not software differences
laced with old usage of current terminology like swap.

i think with NUMA they found they had to keep page reclamation specific to the node zone so kswapd runs as different instance for each node. and of course had to make all the locks not global. i think they also keep scheduling and CPU rebalancing specific to each node.

conceptually seems like a little non-uniformity in memory access goes a short way.
better if they could come up with a cheap small crossbar switch and do away with busses altogether.