jackrabbit-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Ard Schrijvers" <a.schrijv...@hippo.nl>
Subject RE: Performance as compared to simple sql db query is quite bad
Date Fri, 01 Feb 2008 12:17:51 GMT

> Per the suggestion I tried '//*' and it did dramatically 
> well. For 8000 items, first run took ~4000ms and second run 
> ~80ms (probably everything is in cache). 

You still forgot to either set in your workspace.xml
respectDocumentOrder to false, *or* to add an order by clause in your
query. From jr 1.5, default respect document order will be false. I am
convinced your 4000ms will drop pretty steep. Also, iterating over the
8000 nodes will probably be slower then iterating over a resultset from
a database, logically. 


> I am afraid I might not be able to use '//*' everytime, the 
> given example was the simple one, but for each of these type 
> nodes, I would have nodes which carry other information and 
> one referring the other. Doing a  scan everytime and 
> filtering out unwanted would be expensive too.
> Can I specify the schema for the tables, with the current 
> tables with just 2 columns it seems like it won't scale 
> well(storing hierarchy in 2 columns/flat table)
> Ard your documentation will definitely help. Thanks.
> Are there any benchmarks that I can look at?
> -Sateesh.

View raw message