Entry update was issued by Mark Callaghan tests.

XFS file system is known to provide parallelism in operations. There is question-answer text from IRC below, stealing secret knowledge & explanations. Short summary is:

• metadata operations, like modifying ctime, are not concurrent (for the same object)
• direct-io data operations are concurrent (you can ( Read more... )

Мимо пролетал SSD диск OCZ OCZSSD2-1APX250G , я быстренько натравил на него iozone в довольно извращенной форме - 16ть потоков с DIRECT_IO, каждый поток работает с файлом размером в 1гб, размер блока чтения/записи 16кб. Собственно пытался поэмулировать Mysql InnoDB. 2 теста SSD - xfs & ext3 и еще для наглядности обычный SATA и 2.5" SAS на 10к RPM ( Read more... )

long story short:

23:00:29  >>> arekm is deleting 1.5 mln small (few kb) files on xfs... for last 12 hours
23:00:59 < arekm> server is of course doing other stuff, load ~4
23:05:30 < sandeen> arekm, have fun with that :(
23:05:35 < sandeen> reminds me of my lkml maildir cleanup
23:42:27 < arekm> sandeen: why these operations take soo long? ;((
23:42:52 < ( Read more... )

Черт его знает чего там Domas напишет и када, на всякий случай пускай будед тута, комментарии к
xfs allows parallel (and out-of-order, if that matters) DIO, so I got 1500-2700w/s (depending on file size - seek time changes.. :) of random I/O without write-behind caching. There are few outstanding bugs that lock this down back to 250w/s (#xfs@ ( Read more... )

command:
# time bonnie++ -u root -s0 -f -n 256:10000:10000:32
что означает 256*1024 files, размер файла 10000 byte, размазать по 32ум директориям

Results:

type/FS JFS XFS EXT3 ReiserFS real 51m35.716s 98m5.863s 61m17.345s 89m59.596s user 0m2.588s 0m2.920s 0m2.492s 0m2.776s sys 1m45.315s 2m3.608s 1m11.600s 2m13.596s

( Read more... )