Re: Oracle Table Spanning

Serge Rielau wrote:
>
> Doesn't the use of a partitioned table imply that all these reports in
> different formats first need to be cleansed and then copied into the
> partitioned table?
> From how I read this thread it's not obvious that this is an option.
> Partitioned tables deal with scale but not with heterogeneity.
> If copying the data is not an option and it is a requirement to see all
> the data through one interface then I'd guess that some rather complex
> UNION ALL view may be the only option.
> Smells a lot like an information integration problem (just local on one DB).
>
> Cheers
> Serge

I'm still trying to understand the definition of 'Table Spanning', and 'Reports'.

The OP indicated "These reports are unrelational, and very user-unfriendly, and unfortunately, I'm not allowed to manipulate the data." For all I know, the reports could be PDF or images. Nor has the OP indicated whether there is any search or retrieval criteria, possible supporting- or meta- data (although I'd suspect the source from amongst the "thousands of machines" and origination timestampe would be available), or anything else useful. Nor has the OP indicated how the data is to be used, other than having an ASP that access the database.

I'm not even sure whether OP is actually able to define a table to store the data. (Or possibly it's one of those 'table = key + data' dealies.)

Lacking anything more, I'd like to guess that the OP has some form of table, with base supporting data and is either:

1. simply trying to store a large number of discrete 'files'. (This, in turn leads me to thinking Oracle's "Content Management SDK");
2. using the database as a form of queue;
3. doing data collection, as in pharaceutical's test collections; or
4. trolling.

In the first 3 cases, Oracle's partitioning might be used, based on range (date, source, etc.), hash (on the data), or list (source), or even a combination. In this case, the partitioning is simply being used to split the table into multiple tablespaces. Although that does take a bit of DBA effort, there is absolutely no implication on the external code, and no data adjustment required at all. With the big advantage that the partitions can be sliced off, through the transportable tablespace capability, and archived to flopptical or made available using r/o media.

My first guess at 'Table Spanning' is what Oracle called 'partitioned views' back in Oracle 7.3

Just my thoughts
/Hans Received on Sat Jan 24 2004 - 18:32:31 CST