Appropriate Use of Supercolum in Schema

Question

I need to track versions of data in Cassandra. I was thinking that the Version # would be a super column, the entity id would be the row id and then the columns I need to track would be the columns. So, for example, version 1 of entity would be:

• supercolumn: v1
• rowid: UUID
• columns: timestamp, requestid, userid, state, data, etc...

When querying, the most common request will be

• give me version X for record y
• give me the version X - 1 and X + 1 for record y.

Of lesser importance, would be queries such as:

• give me all rows that are in state 1.
• give me all records changed by userid x.
• give me all rows where timestamp is between x and y.

So is the above approach with supercolumns the best approach or is there a better schema to accomplish the above. Are there caveats to using/querying supercolumns that I need to be aware of?

Answer 1

First, do not use supercolumns. They are effectively deprecated and have many drawbacks. In your model you can easily replace your supercolumns with composites by using a composite key with two components, the first being your UUID and the second being your version. All your queries will be supported by this model. Check this out for more information on composites.