Lightbend Podcast: Scaling Akka Cluster to 10000 nodes with Rapid
How Far Can We Scale Akka Cluster?
Recently, Manuel Bernhardt published a lengthy blog post representing months of experimenting and tweaking on AWS. Titled 10000 nodes and beyond with Akka Cluster and Rapid, Manuel recounts this 3-month journey to get Akka Cluster to scale out to 10,000 nodes, breaking the previous limit of 2400 nodes running on Google Compute Engine. The experiment delivered some interesting results, which is why we sat down for this non-expert-asks-questions-to-an-expert session. We discuss:
- How does scalability and clustering work in Akka, from actors on a single machine, to multi-node clusters deployed on Kubernetes and multi-datacenters?
- How does Akka scale differently using Actors compared to the more traditional thread-per-request model?
- What is Rapid, and how does cluster membership work?
- What does Rapid do differently than Akka Cluster?
- What does Rapid bring to the table, and what are the use cases for it?
- Who the heck needs a 10,000 node cluster, and why did you want to scale to that level?
- What’s next for Rapid and integration with Akka Cluster?
Go Reactive with Akka
If you're interested in learning more about how Lightbend can get you set up with Akka Cluster for your distrbuted systems, feel free to schedule a demo with some of our experts!