Ray serve vs Thrift

Is that suitable to use ray serve to replace thrift/protobuf? For example, I would like to provide a service that is not ml-related. Thrift or protobuf is the first choice in my mind. But I wonder whether ray serve can replace thrift/protobuf since ray serve is much more simple.


Any answer for this question? Thanks

Hey @matrixyy! Ray Serve is definitely suitable for this use case :slight_smile:

While Serve has some features that are suitable for ML applications (like request batching), it can absolutely be used for building general services. Like you said, it should be much simpler and easier to get up and running with than building your own services with thrift/protobuf.

@eoakes Could you give a comparison between Ray Serve and Protobuf? Like the advantage and cons. Thanks