ZeroMQ功能

Question

I am looking for solutions for a scenario.

Let's assume a service-oriented architecture (SOA) with hundreds of services. The services are completely isolated – what is behind their APIs is an implementation detail.

Different services can have different security policies – ie who can access the service. For example, a service can be fully public, accessible to a subset of employees, accessible to a subset of other services, etc. Some services may even have that specified on the API level, for example a public service with some internal API calls (is that a bad idea?).

I have touched a bit on ZMQ but not enough to know if this interconnection of services can be accomplished with ZMQ. Any help to decide on whether to continue concentration on ZMQ or not will be highly appreciated.

Answer 1

Are you asking about how to handle security in a SOA? Or are you asking whether or not it is feasible to build a SOA with 0MQ?

The former requires you to build it yourself. You need to define your own security policy between services. Not really 0MQ's domain.

For the latter, yes, 0MQ should allow you to build a SOA architecture. In fact we're doing it right now. Services are encapsulated into containers with a HTTP endpoint handled by nginx, which then reverse proxies the request to a (one or more) nodejs server within through express, which then PUSH messages to workers' PULL sockets on a fair queue basis. Upon finishing processing the request, the worker PUSH its reply back to the server's PULL socket. This way we can spin up or any number of workers we want with minimal disruption to the server. And this is one service.

Service to service communications is handled through REST-over-HTTP.