Apologies in advance if this has been asked before or is a n00b question.
To learn Scala, I'm considering re-implementing a little Java app that uses the SEDA architecture [1], and am not sure if/how I might use actors to implement it. Basically it's a batch application that processes a reasonably large amount of data, and I've broken that processing down into various stages that are quite independent of one another and vary between being I/O bound (both network and disk) and CPU bound.
In the Java app, each stage has a single queue feeding a dedicated thread pool, and each thread pool is sized to try to maximise usage of all three resources (disk, network and CPU). The logic running on each thread within a stage is identical, and there's a clean hand-off between each stage that ensures there's no shared mutable state (either within a single stage or between multiple stages).
Are actors suitable for this kind of thing? Are they re-entrant (so I only need one actor per stage, potentially running concurrently with itself) or can I instantiate multiple actors and hook them all up to a single message queue (so that producers are decoupled from the actor instances)? Is there a different approach that better maps to actor based concurrency?
Previous Topic
