db-derby-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Bryan Pendleton <bpendleton.de...@gmail.com>
Subject Re: Yet another performance question, this time on derby.storage.pageCacheSize
Date Mon, 26 Mar 2012 14:09:16 GMT
>  I forced a garbage collection, and then ran the queries.  The time went down to on the
order of seconds.

This is a fascinating result, though I'm not really sure what it means.

I don't think I've ever given a JVM permission to use 8 GB before.

I'm not sure you can conclude it's directly related to the page cache, though.

If I'm doing the math correctly, only 256MB of that memory was page cache; it's
hard to know what the other 7.75 GB was being used for.

The fact that your system runs just fine on a configuration that's 1/16th the size,
seems to suggest that 95% of the memory in that larger experiment was unnecessary
for your current workload.

It's hard to tell what the system was doing with all those extra resources, though.

But your forced JVM observation indicates that, whatever it was doing, it wasn't
doing it very efficiently :)

With modern VMs you can force a heap dump, and then analyze it later at your leisure.

So you could repeat the experiment (though maybe not going all the way to 8 GB, maybe
just 1GB or 2GB), and then take a dump of that JVM and try to figure out what's
occupying memory, using a heap analysis tool (try the Eclipse memory analyzer, it's
a great tool).

thanks,

bryan

Mime
View raw message