Principles of Data Architecture

Data First

Artifacts must include data flows accurately and track data interactions at a dataset level.

Data is Secure

• Personal and confidential attributes should be encrypted/tokenized
• Data should be categorized based on Enterprise Data Privacy Standards
• Data at rest ( using AWS SSE – KMS ) in motion (TLS 1.2x) must be encrypted

Loosely coupled Data

• Data Exchange Interfaces should be de-coupled
• No Direct connectivity or interfaces to Databases/Application specific data stores
• No Point-to-Point connectivity/interfaces to Databases/Application specific data stores
• Data should be sourced from either an authoritative stores like Electronic Data Interchange(EDI) containers and Enterprise Data Lake (EDL) or It should be from System of Record (SoR)
• Data exchange patterns should adhere to a Canonical Standard ( If applicable)

 Data Should be governed

• Data completeness checks and validation process is implemented and well documented
• Data Quality validation rules should enforced and publish the results periodically.
• Design time metadata and Operational metadata should be tracked to ensure data lineage to support audit requirements
• Strong enforce for Data Stewardship – Stewards are accountable for manage and authority for data

Data flow – Unidirectional

• Data should flow from transactional grain to reporting not vice versa
• Data should not flow from coarse grain aggregators like Reporting System to Transactional systems
• Data should not flow from Operational systems to Transactional systems
• Data should not flow from Enterprise Data Lake (EDL) to Transactional Systems.

Data – Shared Asset

• Datasets are registered and cataloged in centralized system like Collibra system
• All data exchange interfaces should registered through Enterprise process
• Data should hydrated to Enterprise Data lake (EDL) and applicable authoritative data source

Data – Timely accessible

• Real Time or Near Real Time
  • Interface with SoR systems via either Streaming Data Platform or API’s

• Near Real Time and Batch
  • Interface with authoritative data sources via standardized interface patterns

• Access just enough data and just in time

Maintain common vocabulary and Data Definitions

• Dictionary and definitions are published to centralized system like Collibra
• Data should be alignment with Enterprise Data Management (EDM) for well defined, named and consistently throughout the organization.