We have seen CQL 3 add big leaps in Cassandra usability. It is easier to understand for someone moving from SQL development to Cassandra. At the same time, to a Thrift developer, it is again a rework to rewire the map-of-maps schema-free structure of Cassandra to the transposed version of that.

Cassandra 1.2 has been a nice upgrade. It strengthens CQL 3 by adding syntactic sugar to storage techniques to provide collections support. CQL 3 is laconic and supported by all leading programming languages. Apart from CQL, VNodes are the other major development. VNodes is extremely useful for mid-to-large Cassandra setup. VNodes alleviates the load generated by ring alteration by distributing responsibility to stream out across a larger number of nodes and by making the bootstrap process more parallel. A side effect of this load balancing task gets solved automatically. There are other smaller performance boosts for version 1.2.

Cassandra 2.0 is proposed in the latter half of 2013. It will have some improvements, and legacy systems might not be supported. Some of the exciting features of version 2.x are the ability to provide JAR files as triggers, so one can send a registration mail as soon as a new row is inserted in the users table. Compare and Set (CAS) enables clients to update a column by checking if its value matches the provided value. For example, UPDATE loans SET debt = debt + 10000 WHERE user_id = 6f757ca0-dd32-4149-b46d-8f2b744cab02 IF DEBT = 31415.