[Tile-serving] [osm2pgsql] Extremely slow database update (#128)

[sly - sylvain letuffe]

As of my experience, what matters most for diff updates are disk iops. On one single rotating disk, with 32GB of RAM, a 1 hour diff will take between 0.5 and 1 hour (all your I/O will be spent by the update process)
Which mean you better use SSD + 32 GB RAM instead of 96GB and no SSD
If it's too late ;-( try to maximise your iops by building RAID0 arrays* and try to minimise your iops use**.

* instead of a 2 disk raid0 + 1 disk for flat nodes I would try a 3 disk raid0 for all or, if you have other disks (system ? data ?), use them in one big raid5 or raid10 array 

** limit any other disk i/o use on the server. Make use of RAM disk for temporary files (osm diffs, lock files, whatever is in your diff update process)



---
Reply to this email directly or view it on GitHub:
https://github.com/openstreetmap/osm2pgsql/issues/128#issuecomment-39383322
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/tile-serving/attachments/20140402/c160d7af/attachment.html>