[ZODB-Dev] Re: Use of fsync in FileStorage

[Casey Duncan]

On Thu, 29 Jul 2004 23:41:36 -0400
"Tim Peters" <tim at zope.com> wrote:

> [Tim Peters]
> > This commits more-or-less "average size" transactions, but on the
> > high side (about 2KB per transaction record).
> 
> Grr.  That was hallucination, caused by looking at a different Data.fs
> file. It's actually committing transactions smaller than I was aiming
> at, closer to 700 bytes/transaction.  Still, that's thoroughly typical
> if most of your work consists of adding new keys to one IIBTree
> <wink>.

Yeah, I wondered how adding one integer to the BTree (which presumably
mutates one bucket in the common case) resulted in a 2Kb transaction.

I actually tried larger transactions (about 100Kb). As I increased the
transaction size, the difference between having fsync and not shrunk
considerably. This is not too surprising since presumably larger writes
fill the disk cache quicker resulting in more flushing and syncing by
the OS, making ZODB's syncing less aggressive by comparison.

-Casey