Re: [akka-user] Re: Akka-Http perf difference chunked vs non-chunked

Right we have a ticket that sounds rather similar I believe (auto toStricting up unto a given limit), I agree it's a good idea. -- Konrad `ktoso` Malawski Akka @ Lightbend On 3 June 2016 at 11:21:00, daleksan (david.aleksandrow...@gmail.com) wrote: I can

[akka-user] Re: Akka-Http perf difference chunked vs non-chunked

I can confirm that changing the HTTP client to curl does fix the issue. I can see much bigger buffers used in this case. This answers my question however I would say Akka-Http still should be guarded from such case (i.e. never emit such small buffers, due to profound performance degradation, no

[akka-user] Re: Akka-Http perf difference chunked vs non-chunked

I'd recommend you to use wireshark and look at how the JDK http client you are using actually is sending the data, also, let us know if you can get the same result but using other http clients. curl and wget comes to mind. -- Johan On Wednesday, June 1, 2016 at 11:42:04 AM UTC+2, daleksan wrote

Re: [akka-user] Re: Akka-Http perf difference chunked vs non-chunked

I don't see the project attached anywhere - could you put it on github instead? Thanks for prepping a reproducer. -- Konrad `ktoso` Malawski Akka @ Lightbend On 1 June 2016 at 11:05:05, daleksan (david.aleksandrow...@gmail.com) wrote: Hello Johan, thanks

[akka-user] Re: Akka-Http perf difference chunked vs non-chunked

Hello Johan, thanks for your answer. I have a simple project where the issue can be observed. The project does more - it is an experiment to decouple request processing from their acceptance, but it is a small project so running it should be smooth. 1. extract the Gradle project 2. in one term