Back in March a large government health entity approached us regarding the development of a series of new web properties and mobile apps. These “tools” were meant to be used by providers and patients alike for the entity’s context-specific lookup and reference of prescription drugs and national health provider data.
I worked on POC development efforts for what ended up being about forty hours of work spread over 4 weeks. The work involved leveraging Ratpack, a new performance focused, ultra non-blocking JVM web framework.
The REST APIs were meant to run on Heroku with very limited resources and overhead cost, all while providing the mobile and CMS content ingestion teams consistent/reliable performance. The work consisted of four parts:
- The Federal Drug Administration (FDA) National Drug Registry (NDC) REST ETL scripts meant to pick up and consume compressed static files and refresh Mongo-backed storage with updates.
- A reactive, Ratpack-based, REST service leveraging Mongo and Redis for caching with individual requests wrapped within Netflix Hystrix command, fallback, rate limit framework for added resiliency.
- Centers for Medicare and Medicaid Services (CMS) National Provider Identifier (NPI) REST ETL scripts essentially doing the same as their FDA NDC counterparts.
- A reactive, Ratpack-based, REST service leveraging Mongo but without Hystrix integration/command wrapping.
Ratpack’s non-blocking methods and RxJava integration make it fun to play with and undeniably faster than the Servlet-based Apache Sling system I’m used to these days. Using Observable streams definitely requires more thought than just throwing Runnable or Callables at an ExecutorService and waiting for your Futures or callbacks.
The reactive extensions project has frameworks for most languages and this guide is great for understanding how these systems work.
Take a look at the work, tear it apart (#noTests :-()!