Hacker News new | ask | show | jobs
by nedsma 3695 days ago
Totally cool. Alsing and Stannard did a great work on Akka.NET, and I'm happy to see the OP making inroads into Golang projects. How does the message throughput compare between GAM and Akka.NET?

I might as well start contributing to this project.
1 comments
RogerAlsing 3695 days ago
Akka.NET suffers from problems with the Helios transport right now, I don't think the Helios transport do more than a few thousand messages per sec ATM. That being said, there is a new transport coming in Akka.NET 1.5 where we do 100 000 messages per sec in our experiments, so that is a huge improvement.

But that is still more than 8 times slower than GAM.

The reason for Akka.NET (and JVM Akka) having trouble in this area is that every message comes with a sender ActorRef that needs to be resolved. even if the target doesn't touch that ActorRef. The serialization mechanism there is also a lot more complicated.

link
Aaronontheweb 3695 days ago
I've been wondering where you've been lately Roger :p

Releasing Helios 2.0 today, which is a big improvement over the previous version.

Stress test results across 350 connections (ramped up over a 3-5 minutes, each connection doing 250 msg / s): https://gist.github.com/Aaronontheweb/ab3ae953239de81752c357...

Also includes 1 million message throughput tests for duplex and inbound-only receive. On the inbound side we peak out at about 250k / s on a single connection. Write side is slower at the moment due to lack of batching on flush, but that'll be fixed.

link