| Original Post in LinkedIn |
I remember when I started working in IT back in 1995 reporting meant green-bar paper coming out of big machines, and big noise, in cold data centers. If you wanted a new report, it could take weeks. Sometimes months. Then came Data Warehousing; we structured everything. ETL before loading. Star schemas (thanks Ralph Kimball and your Bus Matrix). Governance and Methodology first. It was powerful, clean, reliable, controlled. But rigid.
Then the world exploded with data; we enter to the age of Data Lakes (Big Data here and there). We moved from ETL to ELT. From structured-only to everything: logs, sensors, clicks, documents, images, videos. It felt like freedom. But without governance, many lakes became swamps. The next natural step was inevitable: Data Lakehouse.
We wanted scale and control. Flexibility and trust. Governance embedded into the platform. One foundation for BI, Data Science, and now AI. And today, we are living another evolution emerging. Not pure centralization. Not full decentralization. But Hub & Spoke.
A strong enterprise Hub holding the Lakehouse, Master Data, referentials, enterprise-grade data products. And Spokes (and their tenants), domains, and regions; building their own data marts and domain-specific data products, but aligned to a controlled core of engineering toolkit, naming conventions, data processing frameworks, observability.
To me, this is the next maturity level beyond Data Mesh. It is not about distributing data without control. It is about controlled federation. And in the AI era, that balance matters more than ever.
Comments