how about giving each student(or a group) a huge and dirty dataset; then asking them to cleanse(using Pig+Hive) and and store the data into a datastore(Hbase,Cassandra,.. or even HDFS) for further use(like issuing queries and analysis);

you can add more complexity by specifying how the Hadoop Clusters should be setup, i mean a single machine suffices or they should use a real distributed environment for that;

Thank you for your input. Really appreciated. Can you please provide info on where I can get such datasets?

my suggested (example)datasets to you are:

hope it helped you

Thank you. This will definately help.

Good help.