Am 11.11.2015 um 16:36 schrieb Roberto Spadim:
hm nice, but check my "problem" with one thread per connection, i can think about an estimation with max_connections with thread pool, what i could do? use the same max_connections or use another variable (thread size?), considering that i have a limit of threads running (i'm thinking that this one is better, right?)
the thread pool is typically shorter then max connections, otherwise you won't need a pool and could stay with one thread per connection as before thread pools existed (for most workloads it's better, you need to benchamrk your application) how short or large? depends on how many at the same time active threads and so concurrency you need and at the same time your hardware is capable - too much concurrency quickly ruins performance when the underlying disk layer can't handle it too less concurrency, well, things are slower than they could be and keep in mind concurrency don't make tasks always faster in generel so nobody but you can asnwer your question by trial and measure