DataEngBytes 2023 - MEL-T1-03 - Adric Streatfeild
summary
ANZ builds a cloud data mesh that powers the ANZ Plus app (eg. expense prediction) and claims the first four capabilities cut onboarding from five months to two weeks. Implementation uses the Open Data Mesh spec (opendatamesh.org) with composable JSON/YAML product specs, mesh providers as accelerators, and product patterns like proxy data products and integrated tiers plus T1/single-purpose SLAs. Early lessons: start smaller, build the exact capability consumers ask for, and recruit promoters as advocates.
AI-written summary · about
chapters (17)
- 0:00Data mesh journey intro
- 2:52Why data mesh matters
- 4:19Traditional data challenges
- 7:53Data mesh as microservices
- 9:42Core principles of data mesh
- 11:06Choosing operating model
- 12:35ANZ’s data mesh vision
- 14:03Data mesh as YouTube
- 15:05Building minimum viable mesh
- 17:16First iteration results
- 18:42Data products for ANZ Plus
- 19:47Evolving product specs
- 21:53Types and tiers of products
- 24:01Defining mesh nodes
- 25:28Driving business value
- 26:36Three key lessons
- 29:04Summary
related talks
30mData Mesh – Architecture And Insights From The Self-service Platform Boosting Financial Services
48mBuilding a data mesh
38mData Mesh, Now
30mHow To Select Technology For Data Mesh
43mWhen design meets data: Using Data Mesh for a consumer and business-driven data strategy
44mData Mesh: Lessons from the trenches
27mData Mesh for the Cognitive City
45mRethink Data Accessibility: Data Mesh meets Enterprise Integration