On Thu, 10 Feb 2000, Lawrence Manning wrote:
>Heres the new figures. My poor harddisk must have aged a couple of years
>for all the testing I've done lately.
;)
>Seems 2.2.14 elevator 4 has rewrite cured! Mostly. Dont know what
Yes, as I was saying to Linus the rewrite case gets cured by increasing
the writebomb max-segment logic (from the previous 1 to a 4 tunable
per-blockdevice (or per queue in 2.3.x)) and you now given me the confirm
I touched the right bit. Increasing from 4 to 8 should increase a bit more
performance harming a bit more read latency.
>happened on the 2.3 equivilant run though? I will rerun the test tomorrow
Hug, no idea. Maybe I had a silly issue. Strange. OTOH I mainly focused on
2.2.x for now where I can't certainly break lowlevel drivers and I don't
have the request queue exported to the highlevel layer.
>and if I get different figures (I shouldnt, bonnie seems fairly consistant
>on those 500meg runs.. certainly no more then +/-3%) I will let you know.
Ok. Many thanks for now.
>Machine feels smoothish with the big dd going on. I had a big dd along
>with 4 compiles going and load up near 8 but interactive response was
>still easily tollerable!!! Happy happy... All on one disk of course. I
>also had a mp3 playing (xmms) and it didn't skip once.
Same here of course. The -4 revision for 2.2.x should be the last one
unless strange problem will arise or somebody will suggest better default
values for the three defines I added at the end of include/linux/blkdev.h
(that are per blockdevice (or per queue in 2.3.x) and each blockdevice can
override during the init function.
Andrea
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Tue Feb 15 2000 - 21:00:16 EST