Training algorithms taking 'forever', need some suggestions

I have a training set with 3 input attributes (1 nominal, 2 numeric), 1 nominal output, and 250,000 instances.

I'm experimenting with RandomForests, and LWL mostly, but would like to try some other classifiers.

The training for this takes a very long time, too slow to experiment interactively.

One thing I notice is that weka is not using all 4 cores in my CPU effectively.

What are some suggestions to speed up the training time on the machine I have?

Also, are there any places where I can rent large multi-core machines by the minute (not hour)?

Thanks