[Tile-serving] [osm2pgsql] remarkable allocation time of flat-node cache (#245)

[TobWen]

Tested on osm2pgsql SVN version 0.87.2-dev (64bit id space) on Debian 7 x64 (Wheezy).

The allocation time of the flat-nodes cache is remarkable different on the same system:
* test 1: RAID-0 built with mdadm on two HDDs, filesystem is ext3
`./osm2pgsql --database test --username mapnik --hstore --style /osm/osm2pgsql/default.style --number-processes=6 --slim --flat-nodes test.bin --verbose test.pbf`
running time until nodes gets processed: more than 2 minutes

* test 2: LVM on one HDD, filesystem is ext4
`./osm2pgsql --database test --username mapnik --hstore --style /osm/osm2pgsql/default.style --number-processes=6 --slim --flat-nodes /raid/test.bin --verbose test.pbf`
running time until nodes gets processed: less than 2 seconds

Quick speed test using `dd bs=5M count=1024 if=/dev/zero of=test conv=fdatasync`
* test 1: 160 MB/s
* test 2: 97 MB/s

Anyone with an idea, what's going wrong here? Why does it take such long to create the cache file?
It's repeatable, test.bin isn't existing before and I've tested it with and without filecache.

---
Reply to this email directly or view it on GitHub:
https://github.com/openstreetmap/osm2pgsql/issues/245
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/tile-serving/attachments/20141230/9e5c8430/attachment.html>