Subject re: database replication
Author Dimitar Selensky
Hello Ann,

on 03/07/2001 you wrote:

>I worked for a company that built a cross-database program. We defined the
>core set of actions we required from a database and then built a driver for
>each that implemented those features. It worked pretty well, except for
>transaction management, where the then current version of Sybase bayed at
the
>moon.

I'm just starting to work on a research project (not comercial, it's pure
science ;-)). The basic concept is that we'll need ONE application to
connect SIMULTANEOUSLY to multiple SQL servers, for example to Interbase,
Microsoft SQL server and Oracle. By simultaneously I mean at the SAME moment
of time, i.e. if I run a query, I want all the three servers to execute it
at the same time. I want to get the three results back and compare them.

The goal is to investigate the gain of multiplying the database servers and
to create a fault-free (or close to it) database system. In order to do
this, it'll be best if we can set up a "transparent" layer, that'll receive
queries or just regular SQL statements from a top level application that (in
the perfect case) would know NOTHING that underneath it there's not just
one, but three or even more servers.

Is it possible to write some kind of a driver (even with some limited
capabilities) that'll from one side act as a normal server and from the
other will multiply the SQLs, get the results back and if one of the servers
fails, it'll make a decision based on certain rules and will still return a
result to the top level program?

ANY comments are welcome.

Best regards,
Dimitar Selensky