Data Architecture Best Practices

The problem with data warehousing and lakehousing is that they don’t go far enough. They don’t attack the root input problem: dumb, increasingly siloed and duplicative data and stranded, duplicated logic. It’s the proverbial garbage in, garbage out scenario, unless companies allocate their entire innovation budgets to integration.

More and more apps create the need for more and more reinvent-the-wheel integration, because each app has its own repository and its own data model. This is one reason why big banks have more than 10,000 databases each.

More and more apps with their own repositories is one reason we could be in the Yottabyte Era by 2030. If the world has to store two yottabytes of data per year, 40% of the economy could be dedicated to just storing data–90% of which is duplicated data that’s hard to reuse.

For their part, applications duplicate the core description and predicate logic that should live with the data, where it can be shared and reused repeatedly within a knowledge graph.

A better, data-centric approach puts data and data models first. A FAIR data approach -- smarter data that’s designed to be findable, accessible, interoperable, and reusable–addresses the integration problem up front. No more garbage in, so no more garbage out. Data becomes self-describing with the help of the logic that used to be trapped in applications.

This talk will examine how organizations in various industries use semantic knowledge graphs to solve their data integration and analytics problems at a fraction of the cost of data warehousing or lakehousing. It’s an organic approach to data and logic that can eliminate growing amounts of waste and complexity. Companies can run these systems at low cost in parallel with their legacy environments until they commit to graph data model-driven development, which is when even more substantial benefits will accrue.