Skip to content

Architectural Considerations

johnataylor edited this page Feb 18, 2016 · 2 revisions

We implement stateless services that do the actual work like applying transformations and spinning rules. But the actual metadata that drives the behavior is held in a gist.

The general pattern is we POST data to the service, it fetches the appropriate metadata from a gist, performs the data transformation and returns us back the result.

Using a gist has given us a simple way to edit the metadata. However, the gist has many of the characteristics of a sophisticated metadata store: for example, it is Web-accessible, it incorporates full version control and updates are protected with 2-factor auth.

Because the service is stateless it is easy to scale out, and because the service is stateless the start up time is very low. Overall this is a very cloud-friendly arrangement.

Clone this wiki locally