Cassandra compaction strategy for CRUD transaction

Question

We plan to use Cassandra data store for CRUD type transaction use case. Here is some approx. transaction distribution

Create - 10% of total transactions
Read   - 50% of total transactions
Update - 30% of total transactions
Delete - 10% of total transactions

Given this, is there a possibility of read performance deteriorating over time because of updates/delete/create type transactions. We will plan to use SSDs, we have considerable storage space available. Is there a performance benefit of Leveled Compaction over Size tiered. We plan to use the Size tiered, the default compaction to start with. Please advise.

Thanks Srivatsan

Answer 1

Leveled Compaction Strategy will probably work better for you. Reads will very likely perform better and updates will coalesce and keep obsolete data down. Compactions take more IO, but the IO saved on reads will probably compensate for it, also SSDs reduce concern.