My understanding: contentionless means the RPC being measured uses this connection exclusively, without competition from other RPCs. I think the reference doc also implies that the message is small, so that bandwidth is not an issue. The latency, however, can be affected by many factors -- network speed may not be stable, thread scheduling may fluctuate, so we can assume all RPCs have the same latency. In fact, when we measure performance, no matter it's local or distributed, we always get a distribution. It's a good thing if the deviation is small, but you cannot assume that beforehand.
On Sunday, August 11, 2019 at 11:48:28 PM UTC-7, Robert Qiu wrote: > > Hi There, > > I am new to gRPC and browsing the website of gRPC test. The link: > https://grpc.io/docs/guides/benchmarking/ > > The forgoing sentences confuse me on the website: > > - Contentionless latency - the median and tail response latencies seen > with only 1 client sending a single message at a time using StreamingCall > > > > What is the exact meaning of "Contentionless latency"? Does it mean > that very few data traffic on the network? Something like: only several > messages or packets are sent over the network, the bandwidth of the network > is NOT an issue, the network is NOT a bottleneck. > > If my foregoing understand is right, every RPC call is around the same > latency, right? Why do we need median latency and tail response latency? > These two parameters(median latency and tail response latency) should be > the same value if the network is NOT bottleneck? > > Many thanks in advance! > > > Best， > Robert Q > -- You received this message because you are subscribed to the Google Groups "grpc.io" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/grpc-io/1e9e7b19-62ab-4d91-bb68-59bd4bd45be8%40googlegroups.com.
