Well, I'm thinking in define (maybe via SQL) a set of servers as a
cluster and make the fragmentation rules based on "select clauses",
storing this "configuration" in a specific catalog in global schema.
For example: when a record is inserted in a server which not store
this fragment (no rule matches), it will be automatically moved do the
correct server (the server who matches that record), transparently to
the user. This process involves transaction management and a lot more
validations.
A propose is:
create partition <name> on <site> as <select clause>
This create a replication in this site with the rules of the select clause.
We can provide, for example, users create a foreign key in a site
referencing a table who stored in other(s). In this case, the foreign
key will be created in global schema but the validations of the local
schema's would be managed by DRDBMS.
These are my ideas. Is this supported today?
Gustavo.
P.S.: sorry by the English mistakes...
On 4/19/07, Heikki Linnakangas <heikki@enterprisedb.com> wrote:
> Gustavo Tonini wrote:
> > I'm thinking in project and implement "data fragmentation" based on
> > catalogs inside
> > pgcluster as my university final project. I would like suggestions and
> > would be happy if anyone help me to define it. Sorry if this is the
> > incorrect list...I had no answer on "pgcluster-general"...
>
> This is the correct list.
>
> I'm not sure what you mean by data fragmentation, but we do support
> horizontal partitioning by table constraints. And toasting is like
> vertical partitioning.
>
> What exactly are you thinking of implementing?
>
> --
> Heikki Linnakangas
> EnterpriseDB http://www.enterprisedb.com
>