Artem Pavlenko wrote: > Hi Jukka, >> My interpretation is that this years benchmark was a failure...
> I don't see it this way at all. I think you're missing the whole point > of this benchmark ;) > It was certainly success for Mapnik community as we identified areas > where we can improve and learn from others. Hi, I take my words back, partly ;) Benchmark was indeed a success and all the end users will benefit from it because the next versions of all servers which took part in the competition will have some improvements. It is also possible to get some useful information from the results http://svn.osgeo.org/osgeo/foss4g/benchmarking/scripts/results/benchmark ing2010.odp However, it is important to read all the slides and not just look at the graphs. In the slide #26 (three subsequent Geoserver runs) the slowest, disk-bound run is actually the one that interests me most because that is the situation we have in our production system. It would have been nice to see disk-bound results from all the servers and compare the speed we have now with those. It does not help us a lot to know that we could be 20 times faster if we could keep our aerial photos in memory because we would need not 8 gigabytes but 8 terabytes for that. For comparing our speed with other about similar but most probably well configured systems the 2009 results are still more usable for us http://www.slideshare.net/gatewaygeomatics.com/wms-performance-shootout But as the Constellation team said, good testing is hard. Processors seem to be at the moment so much faster than disk-I/O that a disk-bound test would easily measure just how well different map servers can utilise the disk system used in the test. It will be nice to see what kind of tests we will see next year in Denver. -Jukka Rahkonen- _______________________________________________ Mapnik-users mailing list [email protected] https://lists.berlios.de/mailman/listinfo/mapnik-users
