throughput vs data size vs pipel

mock combine ops, return directly to see max throughput!

this has effect, but not usable! because when increase data size, the time to return becomes long, the throughput will down!

if the execute time of the call is short and return's result is small, the pipeline maybe doesn't need to taken into consideration, but if the call is long and the returned data is large, use multiple thread under the call, and use async multiple calls to pipeline the call, then there will be benefits!