[ZODB-Dev] Re: What makes the ZODB slow?

[Roché Compaan]

On Sun, 2006-07-02 at 17:46 +0100, Chris Withers wrote:
> Yes, indexing seems to be the big problem here...
> I wonder if there's a way to move indexing out of "the zodb".
> The main things I get from ZODB are simplicity of data storage (ie: just 
> playing with python objects) and robustness of data.
> 
> For the first point, indexing is never simple ;-)
> For the second, if indexes become corrupt, they can be rebuilt.
> 
> I wonder if there's something to be said for having a generic object 
> indexing service that didn't use ZODB but used its own local indexes and 
> re-indexed as needed or if the indexes are corrupt?

I've explored a SQLCatalog (all indexes are in SQL) implementation
developed for the ERP5 project a while back. Besides writing a hell of a
lot a SQL, the idea is promising.

-- 
Roché Compaan
Upfront Systems                   http://www.upfrontsystems.co.za