I think it will be much cheaper to rent an instance on AWS or Digital Ocean , on an hourly basis, with Hadoop and Apache Spark already deployed. That way you can play around with much larger datasets.

http://aws.amazon.com/elasticmapreduce/