[OSM-dev] Too many slow queries in db

Brett Henderson brett at bretth.com
Tue Sep 4 10:41:36 BST 2007


Martijn van Oosterhout wrote:
> All the features you list are well supported by postgresql as well, so
> we have choice.
>   
Excellent, I'm pinning my hopes on postgresql, but I know very little 
about it.  Having a play with it is definitely on my todo list.
> Back to the original point: while the OSM database looks but, on the
> scale of biggness it's not really there yet. We're having scaling
> problems but nothing that requires super-duper hardware to deal with.
> Just better software and DB layout. The usual methods for dealing with
> large amounts or data are:
>
> - Partitioning, obviously. I'd suggest dividing the world by
> longitude, say every 30 or 45 degrees. In the future we may want to
> consider latitude splitting but not at this points
> - For our use case, spatial indexes
>   
My hope is that this can be done transparently to code running on top of 
the database vastly simplifying maintenance and operational support.

Yep, longitude is probably idea, in fact something like Oracle allows 
you to define partitions by range which would allow you to define 
partition widths depending on data density within those intervals.
> It's not rocket science.  It just needs to be written... (not
> voluntering just now)
>   
Hehe, me either ;-)






More information about the dev mailing list