I do not understand why people aren't clamouring against postgres's connection model.
I might be missing something but as I understand it, Postgres had chosen to couple the concept of a connection and a unit of concurrency for simplicity of implementation. However this means that even if your server can theoretically handle 5000 concurrent reads you will never get there because opening 5000 connections isn't practical. You'll likely hit a memory limit.
Why is this generally accepted as ok? Decoupling concurrency from connections seems possible via pipelining commands over a single connection. Is it just too late for a project as mature as Postgres?
I might be missing something but as I understand it, Postgres had chosen to couple the concept of a connection and a unit of concurrency for simplicity of implementation. However this means that even if your server can theoretically handle 5000 concurrent reads you will never get there because opening 5000 connections isn't practical. You'll likely hit a memory limit.
Why is this generally accepted as ok? Decoupling concurrency from connections seems possible via pipelining commands over a single connection. Is it just too late for a project as mature as Postgres?