Well, if you can provide us with some benchmarks stating what do you find as 'very slow', maybe we can help you further. But one important thing is to note that a cassandra feature is that as the amount of data in your cluster grows, you are guaranteed (loosely) to get somewhat more constant read/write performance times as compared to a relational db engine.
Have you tried executing the same query in a cassandra-cli/cqlsh, for example? What kind of performance do you get then? If the difference between the cassandra-cli/cqlsh statement and the piece of code you posted is dramatic, that means that it is probably a driver issue, so I would dig deeper in Hector code.