Jeremy Zawodny wrote:
>On Fri, Mar 14, 2003 at 12:03:14PM -0800, Joe Stump wrote:
>>This may help - I just got this from a friend.
>I *really* need to update that. It has become a popular reference.
>But I have some new information to add.
>The short version. If you use LinuxThreads and MySQL 4.0.x (where x
>is > 4 or so), MySQL/FreeBSD is readlly damned good. But using
>LinuxThreads is a key piece there. FreeBSD's threading still sucks
>and can hamper MySQL.
The OP didn't indicate where the files are being kept, and under FreeBSD
this can be significant. I know for example, that using UFS file systems,
are very slow compared to newer file systems. However, I suppose, they
have the trade off that they're perhaps "safer" in the event of a system
crash. But the disk I/O is much much heavier on a UFS file system, than
on some of the newer systems, like ext3 file systems. UFS seems to
insist on doing writes as soon as possible, to prevent accidental loss of
information due to a crash (just my guess, based upon the experience).
I tried to speed up a very slow PostgreSQL database on UFS once.. I
turned off fsync(2) for the database, and that helped some, but the disk
I/O was still enormously extensive for what was going on. The same
database under Linux flies (ext3) (with significantly less disk I/O
I was glad to see the MySQL supports the InnoDB in a partition (raw
disk space). This can eliminate the file system from the equation,
assuming that the necessary caching occurs in either the block device
or in the database engine (using a raw char device).
Warren W. Gay VE3WWG