I keep hearing Sybase IQ works very badly with joins and in my current job main data model (Sybase IQ DB) is View-based with huge denormalized table underneath and still lots of hardcoded logic and number of joins and subqueries (going to 5 levels deep) - all within one huge view.
Am I mistaken, or am I right thinking that classic star schema would still perform better, as number of joins in any query would be limited (instead of computing whole crazy view logic with each query)? I did not have much experience outside Oracle Exadata, but from what I know columnar DBs like Sybase IQ still work fine with joins.
Anyone can advise on good practice around it?