We are using the HTTP/JSON gateway of etcd, not the native gRPC currently. Then, etcd is ultimately a single writer system. Hence, we should verify that we can reach our desired performance goals (expected workloads on expected datasets).
As one element we should write benchmarks for synthetic, repeatable setups and run that as part of our CI (Travis).
We should measure the Python-level ("etcdb" abstraction) performance for random and sequential key read/write workloads.
We are using the HTTP/JSON gateway of etcd, not the native gRPC currently. Then, etcd is ultimately a single writer system. Hence, we should verify that we can reach our desired performance goals (expected workloads on expected datasets).
As one element we should write benchmarks for synthetic, repeatable setups and run that as part of our CI (Travis).
We should measure the Python-level ("etcdb" abstraction) performance for random and sequential key read/write workloads.