A headless data architecture means no longer having to coordinate multiple copies of data and being free to use whatever processing or query engine is most suitable for the job. Here’s how it works.