Enrico Schnepel

Do you have complex requirements for your row level security setup like… • Multiple fact tables containing sensitive data? • Users and or Fact tables belonging to different business domains (Sales, Finance) with partly overlapping access rights? (AD Group “GlobalManagers” can access both, others the one or the other) • Multiple dimensional columns per fact table relevant for row level security? • Column level security dependent on row level security? (Global Data without compensation data, but just for Germany with compensation data) • Complex combinations of OR, AND as well as IN (‘Value’, ‘List’) conditions per user? (Country in (‘Germany’,’Austria’) OR CostCenterGroup = “Sales”) • Exclusions? (Global Finance Data except CostCenter = “CEO”) Did you made more than 2 ticks in the box? This session covers the concepts and implementation ideas to handle these complex Row- and Column-Level-Security requirements at scale in Azure Databricks. In addition to the requirements above, there are some technical restrictions met like • minimal RLS-related overhead for end-user data retrieval (max 5 seconds) and • fast processing time for pre-calculating the mapping tables used during the data retrieval Let’s have fun with building a brain-knot and untie it again.