db-derby-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Army <qoz...@gmail.com>
Subject Re: Optimizer work/issues/discussions....
Date Mon, 02 Oct 2006 17:17:10 GMT
Bryan Pendleton wrote:
> One place I struggle is that I find logged query plans to be rather
> intimidating and hard to read. I've been thinking that it might be
> useful to put some effort into either improving the output of the query
> plan dumper, or perhaps coming up with some other tools for displaying
> and analyzing query selection behaviors.

+1!

> Is there prior art here that we can re-use? Are there good examples of
> tools which analyze and visualize query execution strategies that we
> can learn from and borrow?

I don't know the answers to these questions, but they're great questions and 
hopefully someone else can chime in...

I searched the Derby archives and found an email from a student named Felix 
Beyer about something that sounds promising:

http://www.nabble.com/Feedback-for-extension-tf1967245.html#a5399097
http://www.nabble.com/Explain-Tables-and-Workload-Management-Extensions-tf2009196.html#a5519844

Of particular interest:

<begin quote>

  in a former project I managed to extend Derby to extract the
generated optimized query plans in form of XML files for visualizing
them in an external application. I used the GXL file format for export
and visualized the plans with the JGraph Framework. Internal changes
affected the current Derby structure in two ways: First of all a new
system function was added to toggle query extraction on or off and
second a visitor pattern was used to collect the required information
through a traverse of the query tree after the optimization step.

<end quote>

I don't know the specifics, but that certainly sounds like the right kind of 
thing.  Unfortunately the thread ends with a question from Felix:

"Are Derby Administrators or Application Developers interested in tuning tools 
like this, or could even the developers benefit during testing the optimizer or 
visualizing some internal state of Derby?"

and only one reply:

"IMHO, I'm not so sure how useful this tool would be. I view Derby as a 
lightweight RDBMS. As such, the queries will most likely be straight forward and 
I should be able to figure out the indexes that I'll need to get optimal 
performance."

So perhaps that was the end of that particular missed opportunity.  Unless Felix 
is by chance still following the derby-dev lists...?

Army


Mime
View raw message