I was talking with a friend recently about technology. This individual is a person focused on business intelligence, originally a developer, but now an architect and consultant. They have a fair number of clients and have worked with them to build solutions to assist in analysis and decision-making for all sorts of organizations. This person has primarily worked in the Microsoft stack but has embraced NoSQL, Hadoop, and other technologies. In many ways they view the world as I do, using what works well for a particular situation without prejudice. They want to be effective, using whatever technology may be best in the current situation.
In their career, this person has extensive SQL Server Analysis Services experience and has built many cubes over the years that clients access with any number of front-end tools. I would guess that cube design and construction have made this person a lot of money over the years.
As we talked, I wasn’t surprised to hear my friend say they thought cubes were dead. It was an approach to analysis that they wouldn’t recommend anymore. That is something I’ve felt for some time. As data volumes grow and competition increases, there is a need for more real-time analysis. The processing time for cubes doesn’t make sense.
Hardware advances, query technologies against files in data lakes, and automatic ingestion of large volumes of data into columnar formats have reduced the need for data mart cubes. I see less and less content produced in this area, both by vendors and individuals working with technology. ETL has given way to ELT, and data lakes seem to be far more useful than data marts that pre-aggregate data in predefined ways.
Most of you reading this work in the OLTP space, but there are plenty of you that built BI solutions or interact with those that need them. In the modern, 2020s era, do you find people still building new cubes and taking advantage of ROLAP/MOLAP/HOLAP systems? Or is this now legacy tech you can’t wait to remove from your infrastructure? I think BI is more important than ever, but cubes are dead.