[pgpointcloud] PGPointCloud Performance - what to optimize

[Smith, Michael ERDC-CRREL-NH]

All,

Just starting to test loading Pointcloud data to Postgres (9.2) after
doing a lot with Oracle pointcloud loading and wondering where the
opportunities are for improving performance.

For the following test file
(http://csc.noaa.gov/htdata/lidar1_z/geoid12a/data/1408/20120120_18TWM95520
0.laz), a 7.4M point file, here are my performance numbers for Postgres vs
Oracle:

Oracle (11gR2):
    400   :  1m57.681s
    10k   :  0m42.291s
    100k  :  0m38.627s

Postgres 9.2
    400   :  2m09.042s
    10k   :  1m38.626s
    100k  :  1m51.164s
	
We generally load to Oracle with block sizes of 100K points/per block
(oracle doesn't have a cost penalty for larger block sizes so we tend to
use larger block sizes). For Postgres, the recommendation has been to keep
block sizes to 400. Performance does increase a bit with larger block
sizes but not nearly as much as oracle.

I wondering now, what are the basic things I need to check/optimize for
Postgres ingestion that I'm probably not doing (or is this the currrent
limit on loading performance)?

Attached are my oracle/postgres pipelines for the curious.

Mike

-- 
Michael Smith

US Army Corps
Remote Sensing GIS/Center

  
	

-------------- next part --------------
A non-text attachment was scrubbed...
Name: PostgresPipe.xml
Type: application/xml
Size: 1848 bytes
Desc: PostgresPipe.xml
URL: <http://lists.osgeo.org/pipermail/pgpointcloud/attachments/20130910/4e834d8b/attachment.xml>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: OraclePipe.xml
Type: application/xml
Size: 2129 bytes
Desc: OraclePipe.xml
URL: <http://lists.osgeo.org/pipermail/pgpointcloud/attachments/20130910/4e834d8b/attachment-0001.xml>