Dear Wyo, Graham Jones wrote: > Last time there was serious trouble with XAPI servers I started to think about > setting up my own server (at least for a limited area), but got stuck with the > technology that XAPI uses.
We also looked at XAPI technology and I can confirm that it's a zoo of technologies. And I won't comment the code quality. Returning back to your original question: Do you really need the XAPI? It's currently(?) limited to a small bbox and only one tag predicate and one bbox predicate per query. Our "serious troubles" with XAPI actually have been the reason to build an entirely new solution based on 100% Java technologies: OpenStreetMap-in-a-Box (http://wiki.openstreetmap.org/wiki/OpenStreetMap-in-a-Box ), short osminabox. This solution is based on a standard GIS software stack (GeoServer, PostGIS) and consequently supports the well-known APIs from GIS (WMS, WFS, WCS). Support of XAPI is difficult since osminabox is using a normalized schema and XAPI is based on the non-relational key-value scheme of OSM. If you really want something like a performant XAPI then we have to agree on the use case: If only performance is an issue (given the limited XAPI implementation) then the "cheap" way is to put a cache in front of a local or remote server. DBSlayer seems to work currently only for MySQL. So this is not very helpful here since we need such a connection pooling and proxy layer thing for PostgreSQL. memcached (http://memcached.org/) could be a solution since it's a standalone cache which can optionally informed (e.g. by a db trigger) to outdate the cache so that it get's refreshed. This should bring to probably at most 5 times faster access. Another approach would be to do a complete rewrite of an XAPI implementation :-> Any approach which uses PostGIS - of which currently mainly only osm2pgsql does (and osminabox) - would give you a speedup *and* plenty of query possibilities. Yours, S. 2010/12/13 Philipp Borgers <borg...@mi.fu-berlin.de>: > Which wiki page exactly? I do if I find the page ;) > > I'm highly interested in DB-Clusters. We will play with it in the > future. Suggestions and additional informations are welcomed. I think > osm should use more of these distributed/cluster technologies... > > best regards > philipp > > > Am Montag, den 13.12.2010, 20:18 +0100 schrieb Wyo: >> Philipp Borgers wrote: >> > >> > we use api, xapi, nominatim with a little proxy script written in php. >> > We use memcache for caching. It's far from perfect... ;) >> > >> Nice. For anybody providing their own map, this would be a good proxy script >> instead of the current suggested CGI version. Could you update the proxy >> page in >> the wiki adding your script there? Could you also add a small description >> how to >> install the memcache PECL extension, I guess not everyone is familiar with >> that. >> >> Memcache won't be useable for all the mappers getting their data. IMO we >> need to >> go for a solution which truely can be used as an alternate XAPI server. >> >> Wyo >> >> >> _______________________________________________ >> dev mailing list >> dev@openstreetmap.org >> http://lists.openstreetmap.org/listinfo/dev > > > > _______________________________________________ > dev mailing list > dev@openstreetmap.org > http://lists.openstreetmap.org/listinfo/dev > _______________________________________________ dev mailing list dev@openstreetmap.org http://lists.openstreetmap.org/listinfo/dev
