This is a very simple way to validate the runtime characterstics of streaming subscriptions on the database.
- There are 2 parts to benchmarking: 1) a streaming subscriber, 2) concurrent mutations
- We consider different database models: model_simple.ddl and model_partition.ddl
- We start a streaming subscription and start concurrent mutations (configurable concurrency, default: 1000)
- We check pg_stat_statements to check the max_time of execution of the SQL query. We also monitor the benchmarking app and the database for load (cpu/mem)
- We should test with concurrent subscribers as well.