My buddy Adam put this together for Cassandra. It's always nice seeing somebody do this kind of work and share it publicly.
We do similar kind of benchmarking ourselves for our applications between cloud vendors and instance types (different families) and end up with large spread sheets. In our case we usually equate it to metric per dollar. Such as requests/dollar or GB processed/dollar.
32% performance increase with Oracle JDK 8 over OpenJDK 7. I pity projects tied to this modern day Cobol that have to deal with this crap. Write once, run anywhere (badly, unless Oracle likes your platform enough).
Yes, we were also surprised by the difference. We'd like to try with OpenJDK 8 instead to see if it's performance was better overall, but the charm for Cassandra doesn't include that as an option yet.
3 comments
[ 1.6 ms ] story [ 22.4 ms ] threadWe do similar kind of benchmarking ourselves for our applications between cloud vendors and instance types (different families) and end up with large spread sheets. In our case we usually equate it to metric per dollar. Such as requests/dollar or GB processed/dollar.