There is an explanation of how to lock the JVM into memory here http://www.riptano.com/blog/whats-new-cassandra-065

However from the JVM Heap Size section here 
http://wiki.apache.org/cassandra/MemtableThresholds

For a rough rule of thumb, Cassandra's internal datastructures will require about memtable_throughput_in_mb * 3 * number of hot CFs + 1G + internal caches.

So you have (32 * 3 * 12) = 1152MB for the CF's plus the 1G + internal caches. Running inside of 1 or 1.5 GB is probably not going to work. 

The simple solution may be to reduce the number of CF's.

Looks like 0.7.0 will have per CF thresholds, but you will still need more memory and/or less CF's I think. 

Aaron



On 07 Oct, 2010,at 07:45 AM, Rob Coli <rcoli@digg.com> wrote:

On 10/6/10 9:05 AM, Utku Can TopÁu wrote:
>
> The nodes are still swapping, even though the swappiness is set to zero
> right now. After swapping comes the OOM.

https://issues.apache.org/jira/browse/CASSANDRA-1214

?

=Rob