Slow Geo Queries

Hello,

we are evaluating Elastic Search and have a data set of about 250k
documents, that contain a geo_shape property (mostly polygons). When using
a geo_shape filter my queries become really slow (multiple seconds). I
couldn't really find a way to really change this. I tried quadtrees,
smaller and bigger polygons in queries, etc. No effect. Also I reduced my
query to a match_all query with said filter to be sure nothing interferes.
I was expecting this to go way faster, like other queries, taking a few
milliseconds.

--
You received this message because you are subscribed to the Google Groups "elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

can you elaborate on the shapes a bit. Are you executing any "Star" shaped
poligons on a dense area by any chance?

simon

On Wednesday, May 22, 2013 5:25:32 PM UTC+2, chri...@zoomsquare.com wrote:

Hello,

we are evaluating Elastic Search and have a data set of about 250k
documents, that contain a geo_shape property (mostly polygons). When using
a geo_shape filter my queries become really slow (multiple seconds). I
couldn't really find a way to really change this. I tried quadtrees,
smaller and bigger polygons in queries, etc. No effect. Also I reduced my
query to a match_all query with said filter to be sure nothing interferes.
I was expecting this to go way faster, like other queries, taking a few
milliseconds.

--
You received this message because you are subscribed to the Google Groups "elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

Both the polygons used for querying and in the documents are streets (with
buffers around them), street crossings and districts and other parts of
them. We do a lot intersections using data extracted from place
descriptions in regarding real estates before the result of these
intersections goes into ES. The query then is some district or street for
example, maybe even a whole city.

Am Mittwoch, 22. Mai 2013 21:38:08 UTC+2 schrieb simonw:

can you elaborate on the shapes a bit. Are you executing any "Star" shaped
poligons on a dense area by any chance?

simon

On Wednesday, May 22, 2013 5:25:32 PM UTC+2, chri...@zoomsquare.com wrote:

Hello,

we are evaluating Elastic Search and have a data set of about 250k
documents, that contain a geo_shape property (mostly polygons). When using
a geo_shape filter my queries become really slow (multiple seconds). I
couldn't really find a way to really change this. I tried quadtrees,
smaller and bigger polygons in queries, etc. No effect. Also I reduced my
query to a match_all query with said filter to be sure nothing interferes.
I was expecting this to go way faster, like other queries, taking a few
milliseconds.

--
You received this message because you are subscribed to the Google Groups "elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

I wonder if interpolating the streets and reducing the num of polygon
points make a difference here? How many points do you have?

simon

On Tuesday, May 28, 2013 2:33:47 PM UTC+2, chri...@zoomsquare.com wrote:

Both the polygons used for querying and in the documents are streets (with
buffers around them), street crossings and districts and other parts of
them. We do a lot intersections using data extracted from place
descriptions in regarding real estates before the result of these
intersections goes into ES. The query then is some district or street for
example, maybe even a whole city.

Am Mittwoch, 22. Mai 2013 21:38:08 UTC+2 schrieb simonw:

can you elaborate on the shapes a bit. Are you executing any "Star"
shaped poligons on a dense area by any chance?

simon

On Wednesday, May 22, 2013 5:25:32 PM UTC+2, chri...@zoomsquare.comwrote:

Hello,

we are evaluating Elastic Search and have a data set of about 250k
documents, that contain a geo_shape property (mostly polygons). When using
a geo_shape filter my queries become really slow (multiple seconds). I
couldn't really find a way to really change this. I tried quadtrees,
smaller and bigger polygons in queries, etc. No effect. Also I reduced my
query to a match_all query with said filter to be sure nothing interferes.
I was expecting this to go way faster, like other queries, taking a few
milliseconds.

--
You received this message because you are subscribed to the Google Groups "elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

No, I tried this at least with the polygons for queries and it didn't
really make a difference how big that polygon was. I am not sure, but
since ES uses geohashes, wouldn't it mean that this doesn't really make a
difference?

Changing the precision has a slight effect though, even though it was s
really small one.

Am Dienstag, 28. Mai 2013 14:59:53 UTC+2 schrieb simonw:

I wonder if interpolating the streets and reducing the num of polygon
points make a difference here? How many points do you have?

simon

On Tuesday, May 28, 2013 2:33:47 PM UTC+2, chri...@zoomsquare.com wrote:

Both the polygons used for querying and in the documents are streets
(with buffers around them), street crossings and districts and other parts
of them. We do a lot intersections using data extracted from place
descriptions in regarding real estates before the result of these
intersections goes into ES. The query then is some district or street for
example, maybe even a whole city.

Am Mittwoch, 22. Mai 2013 21:38:08 UTC+2 schrieb simonw:

can you elaborate on the shapes a bit. Are you executing any "Star"
shaped poligons on a dense area by any chance?

simon

On Wednesday, May 22, 2013 5:25:32 PM UTC+2, chri...@zoomsquare.comwrote:

Hello,

we are evaluating Elastic Search and have a data set of about 250k
documents, that contain a geo_shape property (mostly polygons). When using
a geo_shape filter my queries become really slow (multiple seconds). I
couldn't really find a way to really change this. I tried quadtrees,
smaller and bigger polygons in queries, etc. No effect. Also I reduced my
query to a match_all query with said filter to be sure nothing interferes.
I was expecting this to go way faster, like other queries, taking a few
milliseconds.

--
You received this message because you are subscribed to the Google Groups "elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.