Read performance

If a cluster is exhibiting poor read performance, there are some obvious things to check on the application and data model (table definition) before adjusting any configuration settings. As previously mentioned, tables in Cassandra must be designed based on the queries that they are required to serve. If application teams are complaining about operation slowness for multi-key queries or queries that contact multiple nodes, there is little that can be done to improve the situation. On the other hand, if you are confident about the table design and are not violating any known anti-patterns (see the Cassandra anti-patterns section, under Other performance considerations), then configuration in some of these areas may be helpful.

Compaction strategy selection

...