@roji is now writing some benchmarks that don't actually use a database driver as a means to rule out low-level bottlenecks as the root cause for the differences we are seeing with .NET vs. JDBC, Linux vs. Windows and sync vs. async in Npgsql results. We wonder if other teams have done this already, but it seems to have enough value to just do it again. If we are able to confirm the presence of low-level bottlenecks we can pass this test to someone else to do the profiling and analysis.
@roji is now writing some benchmarks that don't actually use a database driver as a means to rule out low-level bottlenecks as the root cause for the differences we are seeing with .NET vs. JDBC, Linux vs. Windows and sync vs. async in Npgsql results. We wonder if other teams have done this already, but it seems to have enough value to just do it again. If we are able to confirm the presence of low-level bottlenecks we can pass this test to someone else to do the profiling and analysis.