[OSM-dev] Towards spatial indexing (was: OSM the mediocre alternative)

A Morris aledmorris2 at gmail.com
Sun Apr 22 11:14:12 BST 2007


On 4/22/07, Robert (Jamie) Munro <rjmunro at arjam.net> wrote:

>
> (see wikipedia about r-trees: http://en.wikipedia.org/wiki/R-tree)



I am the person who originally wrote this wikipedia article - this was one
subject I knew anything about (that had no wikipedia article), having spent
several weeks reading the original paper and implementing it.

In fact the only reason that I implemented the rtree algorithm, was because
the existing spatial search algorithm we were using (in sql server, as I
recall) was

   WHERE Latitude BETWEEN a AND b AND Longitude BETWEEN c AND d.

I think I'm running the risk of preaching to the choir here, but this will
not scale to a large number of users, and it won't scale to a large amount
of data. It certainly will not scale to both.

I would like to do some doing, rather than just talking (actually mostly
reading). So... what I propose to do is to create a test suite that stresses
the database, simulating current and future loads, to enable us to find the
best solution to the spatial indexing problem. (Possibly in the future, it
might help in fixing the data model problems also)

I'm not proposing to define the actual tests, or the actual fixes here and
now - more just to float the idea of a structured and standardised test
suite and check that noone else is doing something similar.

(How long do we have before the current server collapses under the load?)

Cheers, Aled.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/dev/attachments/20070422/03c5a38a/attachment.html>


More information about the dev mailing list