3 ms·

I think the biggest beneficiary of these benchmarks is Postgres. Its drivers are often the bottleneck and everyone who uses Postgres benefits from more work on

I think the biggest beneficiary of these benchmarks is Postgres. Its drivers are often the bottleneck and everyone who uses Postgres benefits from more work on them. And it's also nice to point at TFB when arguing with your architects that 100 rps is not "a lot".

unrealhoang 6y ago
That said, the top C/C++ benchmarks are still using an unofficial fork of libpq client that supports batching. Rust and Java was on the top of the benchmark for a while just because they their own pipeline supported drivers.
- cies 6y ago
  > an unofficial fork of libpq client that supports batching I thought they were just hand optimizing their code to the point they submissions were totally unidiomatic. But now I learn even dependencies get this kind of attention. Years ago I submitted a ticket asking for inclusion of memory usage stats (interesting in small environments) and start-up time stats (interesting in serverless environments). But nothing ever happened on that front; I guess it's because TechEmpower is a Java shop and that are not typical metrics where the JVM shines.