[postgis-users] Spatial Query Benchmarks?

[Craig Miller]

I was only offering the Web App up as an example.  The same approach applies
regardless of the type of application.

- Identify a reasonable set of use-case based scenarios.
- Create a multi-threaded (or multiple single threaded) application that can
execute these scenarios in parallel. Preferably one that is controlled via a
config file.
- Play with the numbers in the config file.  E.g. number of simultaneous
users, size of AOIs, etc.
- Log the time to fulfill each request with 10, 20, 50, 100, 200, 500, 1000
simultaneous users.
- Log any requests that time-out
- Log any requests that are answered incorrectly.
- Mine the data

The program could be as simple as a shell script executing psql.  If there
is a client application, then it is best to write the tests on top of at
least the data access objects portion of the application.  The higher up in
the application you can hook your tests in the better.

How are your users accessing the database?  Psql, c app on libpq, java app
on top of geotools?

--Craig


-----Original Message-----
From: postgis-users-bounces at postgis.refractions.net
[mailto:postgis-users-bounces at postgis.refractions.net] On Behalf Of Ethan
Alpert
Sent: Friday, November 19, 2004 9:59 AM
To: PostGIS Users Discussion
Subject: RE: [postgis-users] Spatial Query Benchmarks?



Well it's not a web app. The problem is it provides users a method of
managing AOI's and querying a fairly large spatial database at user
defined intervals for new intersects of their AOI's. The reason this is
a problem is although the geometry in the database is well defined the
user's AOI's vary in size, vertice count and ring count. So I need to
know at what point will the number of user AOI's affect QoS.

-e

-----Original Message-----
From: postgis-users-bounces at postgis.refractions.net
[mailto:postgis-users-bounces at postgis.refractions.net] On Behalf Of
Craig Miller
Sent: Friday, November 19, 2004 10:53 AM
To: 'PostGIS Users Discussion'
Subject: RE: [postgis-users] Spatial Query Benchmarks?


If you have written an application on top of PostGIS, then the best way
to profile it is to profile the entire solution.  E.g. if it is a web
application then throw WebLoad at it.  By using something like WebLoad,
you can mimic typical use-cases, and then have n number of users
simultaneously hitting it using these typical use-cases.  This will test
the real-world load rather than the theoretical maximums.

--Craig


-----Original Message-----
From: postgis-users-bounces at postgis.refractions.net
[mailto:postgis-users-bounces at postgis.refractions.net] On Behalf Of
Ethan Alpert
Sent: Friday, November 19, 2004 9:36 AM
To: PostGIS Users Discussion
Subject: [postgis-users] Spatial Query Benchmarks?



Anyone have any suggestions for benchmarking? I've got some services
I've built and my management's not going to let me deploy until I can
quantify at what point my it will break on the current architecture.
Anyone thought about how to determine this kind of information? 

How can I know, even heuristically, how many queries/sec my machine can
sustain, given the tuple and/or vertice count in a given table?

Are tuple and vertice count the only things I should be concerned about
or should I also try to thing about geometry shape (convex,concave,
multiple rings,etc) wrt performance.


-e
_______________________________________________
postgis-users mailing list postgis-users at postgis.refractions.net
http://postgis.refractions.net/mailman/listinfo/postgis-users



_______________________________________________
postgis-users mailing list postgis-users at postgis.refractions.net
http://postgis.refractions.net/mailman/listinfo/postgis-users
_______________________________________________
postgis-users mailing list
postgis-users at postgis.refractions.net
http://postgis.refractions.net/mailman/listinfo/postgis-users