In response to a thread, Pat O'Sullivan wanted to know whether it made sense to deploy a physical schema onto NoSQL.
Here is his post:
One additional point/question on the "data modeling for NoSQL"
One big question we keep seeing is "under what circumstances and what business use cases does it make sense to deploy models in a schema-on-write fashion onto NoSQL/Hadoop". In other words while we know that we can technically deploy a physical schema onto NoSQL, is that something that actually makes sense ? Especially when you consider the associated cost of building (and maintaining) the ETL to populate the new schema. The concerns/questions that we see popping up include :
- If you are going to the trouble and cost with such Schema-on-Write structures, then why not do so with a target such as a high-performant column store ?
- Is not such a schema-on-write approach counter to the schema-on-read ethos of many NoSQL/Hadoop deployments?
It feels like this is one area where the dust is still settling a little on what is specific best practice and there seems to be a lot of different opinions from different practitioners out there. I would be interested in hearing other views that folks might have about the dos and don'ts of modeling Schema on Write on NoSQL/Hadoop.
[login to unmask email]
Freelance Production Assistant
Freelance Data, Technology and Science Writer