mmap002 execution time doubled... good or bad sign?
Hi,
I have been running various stress tests on disk for some time. streaming write, copy, read, diff, dbench and mmap002
This is what I have seen:
Quoted 1 line
From 2.4.0 to 2.4.1 with Marcelos patch write were above 10 MB/s
and read >13 MB/s, dbench > 10 MB/s, mmap took around 2m30.
After 2.4.1-pre8 (did not test anything in between)
Write is at 9-10 [lost 1 MB/s] read is down to 11-12 MB/s [lost 2 MB/s] dbench > 9 MB/s [one MB/s there too]
But the really strange one - mmap002 now takes > 4m30 Is this expected / good behaviour? mmap002 abuses mmaps...
/RogerL
Signature · 6 lines
Home page: none currently -- 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 8 lines · Roger Larsson
Hi,
I have been running various stress tests on disk for some time. streaming write, copy, read, diff, dbench and mmap002
This is what I have seen:
Quoted 1 line
From 2.4.0 to 2.4.1 with Marcelos patch write were above 10 MB/s
Quoted 8 lines
and read >13 MB/s, dbench > 10 MB/s, mmap took around 2m30.
After 2.4.1-pre8 (did not test anything in between)
Write is at 9-10 [lost 1 MB/s] read is down to 11-12 MB/s [lost 2 MB/s] dbench > 9 MB/s [one MB/s there too]But the really strange one - mmap002 now takes > 4m30 Is this expected / good behaviour? mmap002 abuses mmaps...
These are probably Jens modifications to the block queuing mechanisms since there were no VM changes from 2.4.1-pre8 to now in Linus tree.
Its not necessarily a bad thing, since Jens modifications make processes throttle on IO sooner.
--
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/