Key Takeaways
Here are key takeaways from the event contributed by Kevin Petrie and Wayne Eckerson in the final session. In some cases, we have referenced the person who made a comment that turned into a key takeaway.
Wayne: People liked the term “data mush” as a the potential downside of data mesh. Wayne hopes it will become a new buzzword!
Kevin: Data product is the primary catalyst for change and adoption of best practices for mesh/fabric implementations.
Wayne: There’s not just one data store, but potentially many stores with an uber store on top. Maybe we call it “data shopping mall”?
Kevin: Data fabric and mesh are symbiotic approaches to managing distributed data—fabric is technical approach and mesh is organizational.
Wayne: Rankin says 90% of the challenge with data mesh is orgnizational and cultural.
Rick Hall: Biggest change over the next 18 months: decentralization of IT and data ownership—and cross-enterprise collaboration to make that feasible
Wayne: Data mesh and data fabric show how to deliver self-service: with organizational, governance, and architectural models – all federated.
Kevin: To succeed, data products must offer views of lineage, usage, and relationships to other products.
Paul Rankin: “Data products demand product management, a discipline many people in data aren’t trained in.
Kaycee: Data fabric has the potential to provide generative AI models the full access, data quality, and explainability they need.
Rick Hall: “without data products, generative AI is just going to be trouble”.
Wayne: Days of IT centralization are over; IT must facilitate, not dictate.
Paul Rankin: Data mesh is good cultural fit for decentralized organization where IT can manage much of the company.
Wayne: Data fabric can inject agility into a monolithic data environment.
Wayne: Roche was impressive in getting out in front of its data mesh with a centralized data platform first.
Wayne: Data quality and data bottlenecks are the KEY pain points that all tools, architectures, methods need to solve.
Matt Fuller: The future of data products is virtual data products.
Sue Tripathi (attendee): Data fabric must align with business strategy.
Kevin: Data fabric supplements data lake & DW.
Phil Hendrix (attendee): Data products should be co-owned & prioritized according to their contribution to the business.
Kaycee Lai: You should be able to discover, build and govern data products in one environment.
Matt Fuller: Data product is the primary catalyst for change and adoption of best practices for mesh/fabric implementations.
Kevin: Data fabric and mesh are symbiotic approaches to managing distributed data—fabric is technical approach and mesh is organizational.
Paul Rankin: Most implementation challenges center on organizational processes and culture.
Kaycee Lai: Catalogs help expert users, but fall short of data fabric/mesh requirements because they don’t help citizen users inspect the data itself.
Rick Hall: To succeed, data products must offer views of lineage, usage, and relationships to other products.
Kaycee Lai: Data fabric has the potential to provide generative AI models the full data access, data quality, and explainability they need.
Kevin: “Without data products, generative AI is just going to be trouble”
Wayne: Data mesh appeals to enterprises whose IT organization is already highly decentralized
Wayne: Data products depend on strong data governance programs to ensure quality
Wayne: Data products require granular metadata to be shoppable, deliverable, and returnable
Matt Fuller: Data lake contributes to fabric/mesh by helping find, integrate, understand, and operationalize data