Write-back/VM question
One of the behaviors of the new VM seems
to be that it starts I/O on a written page fairly early. This "aggressive" write is
great for streaming I/O, but seems to have
a penalty when the application has write-locality. Dbench is a one case, which is write intensive and a lot of the writes are to a previously written page.
I'm not exactly certain why starting write-out
early would cause problems, but I've a couple of
quick questions:
1. Is the page locked during write-out?
2. Is there a tuneable that I can use to
control write-back behaviour?
thanks!
Signature · 8 lines
--------------------------------------------------------------------------
Rajagopal Ananthanarayanan ("ananth")
Member Technical Staff, SGI.
--------------------------------------------------------------------------
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to [email protected]. For more info on Linux MM,
see: http://www.linux.eu.org/Linux-MM/Quoted 11 lines · Rajagopal Ananthanarayanan
One of the behaviors of the new VM seems to be that it starts I/O on a written page fairly early. This "aggressive" write is great for streaming I/O, but seems to have a penalty when the application has write-locality. Dbench is a one case, which is write intensive and a lot of the writes are to a previously written page.
I'm not exactly certain why starting write-out early would cause problems, but I've a couple of quick questions:
1. Is the page locked during write-out?
The buffers on the page are. I'm not sure about the page itself though ...
Quoted 2 lines
2. Is there a tuneable that I can use to
control write-back behaviour?
/proc/sys/vm/bdflush, first value
regards,
Rik
--
"What you're running that piece of shit Gnome?!?!"
-- Miguel de Icaza, UKUUG 2000
http://www.conectiva.com/ http://www.surriel.com/
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux.eu.org/Linux-MM/