Thursday, January 31, 2008
Once You Go Swap, You Can't Stop
How come is it that once LINUX machines start swapping, they don't stop swapping even if they don't need the extra memory anymore?
I have a process that's been running almost 24 hours in real time simply because it slightly exceeded the RAM space available to it. It started swapping and ever since, even though it doesn't need much memory now at all, it's still having kswapd0/kswapd1 invoked and still only using < 10% of the CPU capacity because it's busy thrashing around.
Why does this happen and what can I do to stop the madness!?
*
Meanwhile some links:
A little about big pharma (and the costs of drug development): speaking of which, should I abandon my desire for an academic career, sell out and work for the man? About the subject of the link though -- one thing about drug development is the beginning steps are very much unsuited for a for-profit business as a lot of science is just 1000 monkeys typing at a thousand keyboards (I know I am one ;) ) until one stumbles upon the works of Shakespeare so to speak. Academia and government labs are made for this sort of thing ... industry, not so much.
Howard Dean Daniel Barenboim (don't they look alike?) goes too far here: I don't see where even a de facto one state solution (which is what dual citizenship would be) would guarantee any liberty for either Jews or Palestinians given the track records of the parties involved (especially on the Palestinian side), but he does make some good points and uses the term "leitmotif" ;) ... in general, I think a little bit more of a spirit of hesed would go a long way toward furthering the peace process. What I find odd is that so many so-called liberals refuse to show any hesed or rachmonos toward the Jewish side ... and they wonder why some Jews have grown sour on liberalism?
As my dad always says -- you attract more flies with honey than vinegar.
*
Update -- the process has been running for well over a day now, yet the total CPU time is under 3 hours. Talk about ridiculous!
Update #2 -- it's done! finally! now to see if the results are any good ...
I have a process that's been running almost 24 hours in real time simply because it slightly exceeded the RAM space available to it. It started swapping and ever since, even though it doesn't need much memory now at all, it's still having kswapd0/kswapd1 invoked and still only using < 10% of the CPU capacity because it's busy thrashing around.
Why does this happen and what can I do to stop the madness!?
*
Meanwhile some links:
A little about big pharma (and the costs of drug development): speaking of which, should I abandon my desire for an academic career, sell out and work for the man? About the subject of the link though -- one thing about drug development is the beginning steps are very much unsuited for a for-profit business as a lot of science is just 1000 monkeys typing at a thousand keyboards (I know I am one ;) ) until one stumbles upon the works of Shakespeare so to speak. Academia and government labs are made for this sort of thing ... industry, not so much.
As my dad always says -- you attract more flies with honey than vinegar.
*
Update -- the process has been running for well over a day now, yet the total CPU time is under 3 hours. Talk about ridiculous!
Update #2 -- it's done! finally! now to see if the results are any good ...
# posted by alberich @ 10:56 AM 
