Scientific software services often require long run-times. This is at odds with the synchronous request/response paradigm of the web. Therefore it is necessary to run long-running compute workloads "in the background". This is very often achieved in an ad-hoc manner.
Beanstalk is a simple, fast work queue.
Beanstalk provides a generic interface for (priority) job queuing. There can be multiple independent or collaborating job producers and job consumers written in a variety of programming languages. If a client library for a language does not exist (e.g. R) then implementing the text-based protocol is fairly simple.