Nachtrag zu: [Mapserver-DE] Geschwindigkeit

Arnulf Christl arnulf.christl at ccgis.de
Don Nov 3 11:03:58 CET 2005


Hallo,
auf der EN Seite gabs wieder frisches Futter zu Fragen der optimalen 
Performance. Paul hat es so zusammengefasst:

+++
What happens is that mapserver constructs a SQL query which asks 
(paraphrasing)

"Give me spatial data that is inside this bounding box and has this 
attribute condition"

PostgreSQL takes that SQL and creates an optimal query plan for it. 
Shape file cannot do that.  In theory, PostgreSQL/PostGIS will be 
faster, particularly if both your spatial data and attribute data  have 
indexes.  An optimal query plan will use the attribute index if  it is 
more selective than the spatial index.  The trouble is that  categories 
don't tend to be very selective, in general, so the  practical 
performance of this setup is not going to be super.  If you  goal is 
*only* performance, you are better off simply extracting your  our data 
file into several files, one for each zoom level, each with  only the 
data you want drawn at that zoom level.  That is the  solution that will 
*always* be fastest.

Paul
+++

Es gibt da noch einen offenen Bug in MapServer. Bei rein attributiven 
WFS-Anfragen an große Datenmengen erfolgt der Zugriff ohne 
Umgebungsrechteck. Deshalb greift der GiST Index nicht. Weil da niemand 
mit gerechnet hat startet MapServer dann *zuerst* einen Fulltable Scan 
und setzt *danach* erst den attributiven Filter, das macht u.U. für eine 
Weile die Datenbank platt, also aufpassen.

Gruß, Arnulf.




This site is hosted by Intevation GmbH (Datenschutzerklärung und Impressum | Privacy Policy and Imprint)