COTs/Open-Source ETL tools exist to do this. (Talend, Pentaho, CloverETL, etc.)
With those, you should be able to do this without writing any code.
All of the tools can read from a SQL database. Then you just need to push the data into Cassandra. Many of the ETL tools support web services, which is why I suggested a REST layer for Cassandra might be handy. Using the ETL tool, you could push the data into Cassandra as JSON over REST. (If you want, give Virgil
I haven't tried, but you might also be able to coax the ETL tools to use CQL.
Some of the ETL tools are Map/Reduce friendly (more or less) and can distribute the job over a cluster. But if you have a lot of data, you may also want to look at Pig and/or Map/Reduce directly. If you stage the CSV/JSON file on HDFS, then a simple Map/Reduce job can load the data directly into Cassandra. (using a ColumnFamilyOutput format)
We are solving this problem right now, so I'll report back.
---- Brian O'Neill
Lead Architect, Software Development
Health Market Science | 2700 Horizon Drive | King of Prussia, PA 19406