I don´t know the internals of a commit operation, but is it composed
only of a write operation? The are many tests "reads" that could be
done in paralel, leaving only the "write" part do be serialized. Am I
wrong?
2006/5/22, Garrett Rooney <rooneg@electricjellyfish.net>:
> On 5/22/06, Ben Collins-Sussman <sussman@red-bean.com> wrote:
> > On 5/22/06, Garrett Rooney <rooneg@electricjellyfish.net> wrote:
> >
> > > Making things work concurrently might be complex, but at the very
> > > least it seems like we should think about adding some locking... If
> > > you're going to append to the prototype rev file you lock it (both
> > > file based and a mutex if you're a multithreaded process, just like we
> > > do for the repos lock), same for the changes file, etc.
> >
> > Yup, this would make concurrent writes to the same txn safe. But it
> > also means that FSFS would never get any of the benefits of
> > parallel-processed commits. There might be multiple httpds all trying
> > to write at once, but then serialized at the fs level.
>
> Only so long as the actual write is occurring. I imagine there's
> still gains to be had for reading network traffic concurrently,
> preparing to do the write, etc. Not saying it couldn't be made
> better, just that a "slow, but safe" approach seems best at the
> beginning.
>
> -garrett
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
> For additional commands, e-mail: dev-help@subversion.tigris.org
>
>
Received on Tue May 23 02:43:48 2006