The concept-oriented model tries to overcome some problems of RM by relying on two constructs: sets and functions. In contrast, RM uses only sets. The idea is that data can be stored in functions and transformed via operations with functions.
Except when you have large, telemetry style datasets e.g. web/product analytics which won't fit. Or when you are trying to build a wide table and you run out of columns. Or when your favourite SaaS products gives you highly nested JSON data.
RDBMS works great up until the point that it doesn't.
> Except when you have large, telemetry style datasets e.g. web/product analytics which won't fit.
Web analytics was one of the first applications for Greenplum. My understanding is that Yahoo collected tens of billions of events per day in the mid-2000s.
> Or when you are trying to build a wide table and you run out of columns.
HAWQ can run SQL queries over Hadoop clusters. Clickhouse's table width is limited by how much RAM you give it.
> Or when your favourite SaaS products gives you highly nested JSON data.
This is why major databases have JSON querying capabilities and why it's been added to the next SQL standard. PostgreSQL even allows you to define indices on fields inside your JSON structures.
Better yet: decompose the highly nested data. Relational databases begin to shine when you get past at least first normal form.
> RDBMS works great up until the point that it doesn't.
RDBMSes do work great until they don't. Which means they are almost always the best solution and almost always remain so.
Folks regularly overestimate the size of their problem and underestimate the capabilities of the literally dozens of RDBMSes now available for use. Yes, it irks me.
Disclosure: I work for VMware, which sponsors Greenplum development.
RDBMS doesn't work but columnar databases such as BigQuery, Snowflake, and Presto make it work. I agree that web/product analytics data is hard to get it done the right way thus people rely on specialized solutions such as Mixpanel and Amplitude but I believe that as companies started to be more privacy-concerned and want to fill the gap in between the product and other company data such as marketing, finance, etc, it will change over time.
See how it's possible to run funnel, retention, segmentation queries in Firebase Analytics data on BigQuery with SQL: https://www.youtube.com/watch?v=7tXVG785_0M
