On Fri, 21 Jun 2002, Jeff MacDonald wrote:
> Eekk.. good point. back to the drawing board..
>
> Could we get a list of folks that would be willing to run a replicated
> server,
> make them into a "server pool" and get the mirrors to connect to them..
>
> I'm not sure if that would be a great solution, cause i'm not sure if most
> of the traffic is because of images, static code, or whatever would be
> coming
> from the database...
>
> I guess at the very least it's still good for discussion.
It would also be a great test case and learning experience of distributed
databases and replication.
I would be very interested in further discussion along these lines.
I don't know about the existing site but perhaps we could split the
mirrors on DB mirrors and www mirrors. The list visible to users will only
list www mirrors.
We could have two DNS entries. One fixed to the main DB site and another
to a round robbin list of DB mirrors. The web site programs could try to
connect to the round robbin DBs first and if that fails then go to the
main site.
We could even have configuration files so mirror sites could
perhaps specify a preferred DB mirror and then the site would try the
preferred first, then the round robbin and then the primary site.