We're having a good time with Cassandra on AWS ingesting more than 200 GiB per day uncompressed. I don't know how you're running your IOPS numbers, but consider allocating large GP2 EBS volumes rather than PIOPS--you'll get a high baseline for not that much money. The provisos you'll see about knowing how you expect to read before you start writing are absolutely true. :)

(Hope that might be helpful! A bunch of us hang out on IRC at #cassandra if you're curious.)