1) It's rare to have enough insight into the internals of a particular datastore to accurately predict how it will perform on a particular workload. Whenever possible, early testing on production-scale workloads is essential for planning and proofs of concept.
2) Database capabilities are a moving target. E.g., the performance improvements to pgsql's LISTEN/NOTIFY are essential to its ability to handle this particular workload. In previous jobs, I've had coworkers cite failed experiences with 15-20 year-old databases as reasons for not considering them for new projects. Database tech has come a long way in that time.
3) Carefully-tuned RDBMSs are more capable than many tend to admit.
1) It's rare to have enough insight into the internals of a particular datastore to accurately predict how it will perform on a particular workload. Whenever possible, early testing on production-scale workloads is essential for planning and proofs of concept.
2) Database capabilities are a moving target. E.g., the performance improvements to pgsql's LISTEN/NOTIFY are essential to its ability to handle this particular workload. In previous jobs, I've had coworkers cite failed experiences with 15-20 year-old databases as reasons for not considering them for new projects. Database tech has come a long way in that time.
3) Carefully-tuned RDBMSs are more capable than many tend to admit.