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.
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.
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.
Comments
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.
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.
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.