I have case that needs manipulate large stream of JSON and injecting it to Apache HBase. Our system works on Node.js with Mongo then, since we needs to enhance performance so HBase is choosen to handling the big data things.
To enchance my system scalability, I prefer using Actor Model by Akka for messaging instead any other messaging queue system. It's because Actor Model that Akka provide gives me any advantages about fail safe, Actor management, and other features that's very helpful to make my job done. But it still in JVM layer that directly injecting and consuming data from HBase.
I want my Node.js apps also works under Akka system maybe using node-java. Is it good practice? If it's not, is there any solution that Node.js can communicate with Akka?
ps. my question here is about how to works with Akka and Node.js, not arguing about "why choose to use Node.js when JVM has really fast JSON manipulating library?", it's because our system are already benchmarked and Node.js was the choosen one to handling JSON manipulation. Also it's already on production stage, so migration totally from Node.js to full Scala is not our priority today.
Just to clarify, Akka implements message passing as it's concurrency model and it supports Message Queue patterns (e.g. broadcast, pub-sub). However, you'd be better off looking at MQ solutions if that is really what you need.
I think going down the path you proposed (running NodeJs with Java interop) will yield little benefit whilst adding significant complexity for the long term.
Better to look for an answer from an architectural point-of-view.
If I had to decide, I would create a Scala / Java Akka microservice that sits between your NodeJs front-end and HBase. You can get a quick Proof of Concept running (which you can back out on relatively easy).
PS. If you are committing yourself to HBase, I would highly recommend you to also look into Apache Spark, which makes taming Big Data easier.