On Mon, 2006-10-09 at 14:13 -0700, Bruce Dembecki wrote:
> On Oct 9, 2006, at 7:15 AM, Ow Mun Heng wrote:
>
> > Hi All,
> >
> > Just wanted to know if it would be faster/better to implement this
> > option into my.cnf
> >
> > innodb_file_per_table = 1
> >
> > which would essentially make each table a file on it's own rather than
> > have it all in 1 file.
> > My belief is that it would be slightly more advantageous compared to 1
> > BIG file.
> >
> > eg: 1 10GB file would perform poorer than 10 1GB files.
> >
> > Is this statement true and how far is is true?
> >
> >
> There are some minor performance benefits here when run against
> of course every little bit helps.
Of course.. Esp when the Box is no Big Iron. :-)
> corrupt InnoDB table file under file_per_table means only one table
> is at risk as opposed to the entire database.
Didn't see it that way. That's Good as well.
>
> One of the big things that really really helps us is having files
> that are appropriate in size for the data... That means mostly our
> ... big performance gain when moving files
> around the network.
This is useful as well, for backup purposes.
> Files can still be stored on different storage devices by making
> symlinks within the data directory for specific database directories
> or even specific table files.
I thought symlinks were only applicable when doing/using myISAM tables?
Didn't I read somewhere that for innodb, you can't really copy the files
to another directory (I presume this meant symlinking as well) because
they're still tied down to the BIG Tablespace/Table Log File?
>
> We really really like innodb_file_per_table - but mostly because it
> makes our lives easier in many ways, not so much for performance
> reasons.
That's more than enough reason for me to switch.
Thanks for all the answers.
