Re: [Gluster-devel] AFR between two bricks over 3000 miles

[Onyx]

Wouldn't a fairly easy solution for slow- and high latency links be to 
have a large cache on a local filesystem? Like there is now a cache in 
memory, make a cache definable on the local system/site?

Anand Avati wrote:
> > Bandwidth is so cost effective now days I am not worried about slow links.
> > I can't however with any amount of money speed up the speed of light in a
> > glass fiber. : ) I understand their is not immediate fix, however do you
> > have any rough estimates? Are we talking weeks, months or years?
> >     
>
>
> AFR currently takes a 'synchronous' approach over its subvolumes.. i.e.
> completes operations over all its subvolumes before returning. Essentially
> it works as fast as the slowest subvolume. Changing this to work as fast as
> the 'fastest subvolume' needs some major changes (handling some race
> conditions) which are too disruptive to be introduced in 1.3.8. We are
> considering bringing this 'optimistic' behaviour (as opposed to the current
> 'pessimistic' one) for 1.4. That is about 2 months away.
>
> avati
> _______________________________________________
> Gluster-devel mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/gluster-devel
>