node.js vs java - compute intensive computation -

I am currently evaluating a backend architecture for an API, it looks deeper in node.js and With this JSB the J2ee application server war files API will have to calculate considerable CPU-intensive tasks - for training of more specialized neural networks. Do you have any experience related to such setup and / or what package should the machine use to learn? Should any other API technologies be considered - for example Java Micro Service?

Thanks a long way ago, Philippe

I have at least two subtle services The recommendation to accept an API request is to calculate the other (s) you can choose any language for both of these. For example:

1. A node.JS API service - HTTP receives trustworthy requests The service calls another service such as Transactional RPC Protocol.
2. A stateless C ++ / Java / Go service - receives count tasks through DND RPC, calculates, returns results. As a result, you will offload Frontend Service, it will not be on high load. The second service is written through CPU optimized library / language. Besides, you can scale the second service in case of too much load.