On Fri, 2005-06-10 at 07:37 -0500, David Johnson wrote:

> If a distributed store-and-forward mechanism is required ...
>
> Run a set of triggers that inserts key information for new/updated rows
> into a submission queue table, ordered by timestamp (need to use
> millisecond or finer granularity).
>
> Run a parallel task that processes the data from the submission queue,
> and forwards the rows from the database back to the central server ...
> in near-real-time.
>
> Since the information is stored in a table, if there is an outage
> anywhere, no data will be lost. Transmission will resume once
> everything is working, and systems will sync up.

Thanks, we will probably end up doing that. That is how a lot of other
data gets moved around our databases already.

>
>
>
>
> If centralization is really the goal, and you can afford to have remote
> sites shut down if there is a problem at the central site or the
> communications, you might be able to get a larger box and run Classic.
>
> Does anyone know if the issues with the 64 bit compiles were resolved?
> If so, a big opteron or athlon X2 type system might be worth examining
> (depending on your budget). Toshiba has announced plans to build
> mainframe class systems around the Itanium, which will run Windoze or
> linux.

I like to avoid centralizing too many client on 1 machine. I like to
get a bunch of cheap rack mounted Dell boxes, slap linux on them and
roll them out. If only Firebird could cluster like Oracle RAC (at least
like they claim to be able to) then I could just add more processing
power whenever I needed it by adding another box to the cluster...
Wishful thinking for now because there is no way I am going to pay that
much just to use their software (not to mention hire a dedicated DBA).

Thanks for your suggestions.

-steve


[Non-text portions of this message have been removed]