Maximizing Throughput and Scalability for Akka Streams
summary
Shows that inserting async boundaries or using mapAsync parallelism turns fused single-actor stages into real parallel work, e.g. CPU-bound spin loops (~10ms) scale with cores (1→4 gives ~4x), while nonblocking calls with parallelism=1000 cut a 50–60s run to ~100ms (≈1000x). Reminds that actor mailboxes give a default 16-element buffer, that unordered mapAsync avoids head‑of‑line blocking, and that external partitioning or cluster sharding (or scaling pods/consumer groups) provides scalability and HA.
AI-written summary · about
chapters (5)
- 0:00Introduction to Akka Streams
- 3:36Maximizing throughput for single stream
- 16:59Partitioning streams for throughput
- 24:46Partitioning streams for scalability
- 28:39Key takeaways and Q&A
related talks
54mScala concurrency patterns with Actors
48mStreaming all the things with Akka Streams
46mTame Your Data with Akka Streams
29mTHINGS I WISH I KNEW WHEN I STARTED WITH AKKA
01h 14mAkka.NET: The Future of Distributed Programming in .NET
01h 07mA Practical Introduction to Akka Streams
11mStreaming Libraries Shootout
01h 32mFrom Spring + Java to Spring + Akka - A Journey of Discovery