access paths. Example: lots of secondary indexes. • The application depends on identifying rows with sequential values. MySQL autoincrement or Oracle sequences. • Cassandra does not do ACID. LSD, Sulphuric or any other kind. If you think you need it go elsewhere. Many times people think they do need it when they don’t. • Aggregates: Cassandra does not support aggregates, if you need to do a lot of them, think another database. • Joins: You many be able to data model yourself out of this one, but take care. • Locks: Honestly, Cassandra does not support locking. There is a good reason for this. Don’t try to implement them yourself. I have seen the end result of people trying to do locks using Cassandra and the results were not pretty. • Updates: Cassandra is very good at writes, okay with reads. Updates and deletes are implemented as special cases of writes and that has consequences that are not immediately obvious. • Transactions: CQL has no begin/commit transaction syntax. If you think you need it then Cassandra is a poor choice for you. Don’t try to simulate it. The results won’t be pretty.