What I have noticed is that it is directly the number of concurrent requests I am sending over the given connection at any time e.g. if I am using 2000 connections to send simulated load on a single connection I see crazy high latency - however if I (for testing purpose) send those across a small pool of those connections (e.g. 32) and force the requests to block, the latency is more like 1-2ms (though the number of connections in the pool is too low and it ends up being the bottleneck - taking on the previously observed latency).
On Wednesday, September 14, 2016 at 1:25:06 PM UTC-4, [email protected] wrote: > > I am trying to send many requests over gRpc, however am running into very > poor performance under simulated load via. wrk with a 100ms rpc request > deadline. > > - ubuntu 14.04 x86 > - go1.6.2 > - 1 Connection being handled via. gRpc > - The operation being performed on the server end is always less than 1ms > - gRpc time is anywhere from 30-500ms > - this test is performed with both client / server on the same machine > > I attempted a loosely handled open / close connection method which > appeared to give a more stable overall time (3-150ms). > > I am aiming for sub 100ms at all times. > > What is the best practice for handling this? > -- 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 post to this group, send email to [email protected]. Visit this group at https://groups.google.com/group/grpc-io. To view this discussion on the web visit https://groups.google.com/d/msgid/grpc-io/44eb814c-bb9d-4267-bb2f-cb070418ba40%40googlegroups.com. For more options, visit https://groups.google.com/d/optout.
