[Qgis-psc] Supporting a permanent performance testing infrastructure for QGIS server

Régis Haubourg regis.haubourg at oslandia.com
Fri May 4 07:58:09 PDT 2018


Hi dear and beloved PSC,

At Oslandia we may soon have the opportunity to build a new infrastructure for QGIS server performance benchmark.

We in fact already started to build a light weight python tool that can be wired in a continuous integration environment
like gitlab CI or Travis. It can produce reports and graphs, and inject new performance tests easily using a simple yml
config [0].  We'll start comparing QGIS LTR, Release and master, and inject various use cases progressively. 

For the sake of our client needs, we would rent a dedicated server so that we can run tests on isolated machines that
will not be influenced by external bias - at least as much as possible. The rent will run only for the duration of our
project.
Our customer, as much as ourselves, would really like to open that project and make it live on the long term besides OGC
compliance tests and all current continuous integration so that the community can use it and enrich it with new test
cases.

A dedicated server is very much needed for performance benchmarks, and this costs something between 900 euros and 1200
euros a year - no including the maintenance of the test framework.   

We would like to discuss if we could find a way to fund or compensate in some way the costs of that permanent
infrastructure rental and maintenance.

Among the ideas we have, we see mainly two possible options in the current situation

- Directly fund that on behalf of QGIS.org 
- Oslandia funds that directly and we deal the equivalent sponsorship level. It could probably be something like Silver
or above depending how much support and maintenance we provide for adding new tests. 


We don't think grant applications are fitted for long term funding, but depending on the interest we could apply for a
grant that would improve the performance logging status in QGIS and add some critical test cases (like provider
comparison for instance)

How does the PSC feels with that proposal ? 


Regards, 
Régis

[0] http://hekla.oslandia.net/pbl/graffiti/report.html



More information about the Qgis-psc mailing list