Suppose I have a Cassandra table like this:
CREATE TABLE keyspace.last_modified ( category text, subcategory text, id text, created_at bigint, PRIMARY KEY ((category, subcategory, id)) )
- The primary keys are used deliberately to make each row to be overwritten
- This table is used deliberately to be consumed by two materialized views, with created_at as the first clustering key.
The questions are:
- Is Leveled Compaction Strategy (LCS) really the most suitable compaction strategy for overwritable data?
- Should the materialized views also use LCS because they are also mutable?
- When should we concerned with compaction strategies? How to profile compaction strategies?
Currently, I already selected LCS for these, but my data is not too big yet.