Demonstrating hot_standby overhead
A few days ago I went to Pilsen, to talk about database replication (general theory and some details about PostgreSQL replication) at a local JUG meeting. One of the questions you usually get after such talk is what is the overhead of a hot standby. The usual answer is "minimal" but that's not exactly a reliable answer ... so let's see a bit more thorough answer based on a simple benchmark.
