Managed database services solve a real problem well: operational overhead. Automated backups, point-in-time recovery, failover, patching — these are genuinely hard to operate correctly, and managed services handle them reliably. The tradeoff is performance ceiling and configurability. That tradeoff is invisible at small scale and painful at large scale.
The parameter problem
A self-managed Postgres instance gives you complete control over hundreds of configuration parameters: work_mem, effective_cache_size, max_parallel_workers, random_page_cost, checkpoint_completion_target, and hundreds more. These parameters interact with each other, with your hardware, and with your workload in ways that are specific to your situation. The optimal values for an OLAP workload on NVMe storage are completely different from the optimal values for an OLTP workload on network-attached storage.
Managed services offer a small subset of these parameters through a configuration UI. The ones they expose are the safe ones. The ones that actually matter for extreme performance tuning are locked — set by the platform to reasonable defaults that are optimal for no one in particular.
The query planner ceiling
Beyond configuration, managed services can't let you replace the query planner or add custom execution hooks. This is where performance engineers hit the hard ceiling: they've tuned everything accessible, their indexes are optimal, their configuration is maxed out, and the planner is still choosing bad plans because the statistics model doesn't capture their data distribution accurately. At this point, on a managed service, you're stuck. The only option is to rewrite queries to force better plans, and that's a maintenance burden that compounds over time.
What Dreambase adds
Dreambase works as a layer above managed services: it adds query intelligence and adaptive optimization without requiring access to the underlying engine configuration. You keep the operational benefits of the managed service. You gain the performance benefits of an optimization layer that understands your specific query patterns and data distribution. The ceiling moves.