[OSM-dev] what server next?
grahamjones139 at googlemail.com
Tue Sep 8 19:41:38 BST 2009
I would like to look into a distributed version of XAPI - I agree with Tom
that the current XAPI server is not ideal (but this may be because I can't
work out how it works).
I envisage an XAPI server utilising a 'standard' PostgreSQL database
produced by osmosis, kept up to date by the daily/hourly diffs etc, along
with a simple web server front-end to do the XAPI to SQL translation.
I made a start coding a simply python based server to do this, but haven't
got very far (made the mistake of deciding not to write my own parser but
use a parser generator - terrible mistake!). Maybe I should give up and
write it in C - I'm sure the production version will be in C for speed
This would work for a single 'main' server, but I like the idea of it being
distributed with lots of little ones (for example the computer in my attic
could serve Northern England, someone else could do Belgium etc.). I don't
know how to deal with re-directing the requests without a central main
server though...any ideas?
Having a nice fast XAPI system would make more dynamic maps (e.g. a vector
layer of POIs obtained from XAPI over a static mapnik background) a bit more
usable - all of the examples I have seen so far are a bit sluggish.
2009/9/8 Tom Hughes <tom at compton.nu>
> On 08/09/09 08:02, Patrick Petschge Kilian wrote:
> > The first one would be a (decently fast) OSMXAPI server. Since the 0.6
> > switch there seems to be a shortage of XAPI servers. If there was a
> > stable, fast and up to date XAPI server it would help lots of people and
> > it might reduce load on the main API.
> We've already given XAPI a server, and before we give it any more
> hardware I would need some serious convincing that XAPI as it currently
> exists is actually workable and scalable in some sensible way because as
> far as I can tell at the moment it isn't.
> As I understand it the database takes so long to load that if there are
> any working servers they are way out of date, and even when a server is
> working it can only serve a couple of users at a time which, when a XAPI
> query can often take minutes or hours to run is clearly not practical
> for serving a large community.
> There are also issues with the run time that the code uses - it's
> horrible ancient and crufty and requires various kernel security
> features to be turned off.
> Tom Hughes (tom at compton.nu)
> dev mailing list
> dev at openstreetmap.org
Dr. Graham Jones
email: grahamjones139 at gmail.com
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the dev