Hi I've been trying to compare Berkeley and eXist performance. Im doing this through PHP and specific queries because thats what I need for my project, but even though the odds are against it eXist seems faster!?

eXist (java, no JIT) is running on Apache Tomcat and PHP 5 communicates using SOAP (all on same machine) using a very beta 3rd party PHP module. Berkeley is called using their own PHP module. A typical result from my test script might look something like this:

(A) Berkely DBXML. returned 18 articles
query:let $con := collection("test.dbxml")
for $x in $con/metadata
where $x/labels/s="bob"
return $x/id

(B) eXist XML DB. returned 18 articles
query:for $x in /metadata
where $x/labels/s="bob"
return $x/id

Timings: A:0.2554 sec, B:0.245 sec, Total:0.5004

Looking for <s>bob<b> in a <labels> tag which may have several other <s>, Im using indexing, does anyone have any ideas whats going on and if I can improve speeds?

Actually nevermind, I tried a more challenging query and it really sorted the men from the boys! Average results: Berkely 0.26 seconds, eXist 5 seconds

(A) Berkely DBXML. returned 129 articles
query:let $con := collection("test.dbxml")
for $x in $con/metadata
where $x/labels/s="jim" and $x/labels/s="bob"
return $x/id\n\n
(B) eXist XML DB. returned 129 articles
query:for $x in /metadata
where $x/labels/s="jim" and $x/labels/s="bob"
return $x/id

Timings: A:0.2609 sec, B:5.0714 sec, Total:5.3323

Hello,

There are various configuration/coding things that could cause
your results.
This discussion could get interactive. Let's take it off the group.
Please respond directly to gmf (AT) sleepycat (DOT) com.

A few questions:
1. Are you creating a new XmlManager and opening the container
on each request? If so, you should not?
2. Are you measuring the results from a number of queries (i.e.
getting a caching effect)? If the answer to 1 was "yes" then
you are not.
3. Are you sure your indexes are being used?
4. How large is your container?

Regards,

George


tfga wrote: