But I still assume these are legitimate conclusions that each present what we can work on to improve and see how the protocol handles the stresses. At this take a look at measurement, we are able to start to see the tail-finish latency begin to enhance while maintaining a comparatively consistent average across the board. These checks had been chosen to check a good spread of outcomes with out attempting to actually benchmark PlanetScale and the underlying mysqld processes we use. This check consequence genuinely surprised me in each eventualities. What this does prove is that we aren’t adding something measurably worse to the connection within the fastest situations. Whereas this is just one experiment specializing in latency and comparing the protocols, there are different benefits that aren’t discussed here that come with an HTTP API that I will probably be speaking about once we start to publicly doc these in the approaching months. Our new APIs aren’t just gRPC.
