it uses lightweight Fibers instead of Threads to handle concurrency, so it allows more db operations per process
Removed unused middlewares
Removed prometheus http collector, it adds around 20% and not used in node, send a patch to the prom client with improvement https://github.com/prometheus/client_ruby/pull/316, I guess it can be improved further
Final results for create endpoint, it is not 10x, only 3x RPS and 8x better latency
Yes,
Final results for create endpoint, it is not 10x, only 3x RPS and 8x better latency