Premise
A longtime client of Axis Group retained my team to explore Microsoft Fabric and evaluate its usefulness for a small enterprise. At the time Microsoft Fabric was under 2 years old and had many features in alpha and beta stages. My team was tasked to evaluate:
- Data ingestion
- Structured ETL in a data lake pattern
- Data enrichment
- PowerBI recreation of existing production reports
Our evaluation was to be presented to the client as both a report and a functional system ready to be handed over to key stakeholders. Internally myself and the senior project architect were tasked to collect information about Fabric for internal evaluation. In this project my role was to design, implement and test the data architecture in collaboration with a senior architect.
Process
I began the engagement by conducting user interviews with client data subject matter experts. The client used ServiceNow to log maintenance requests, system status and incident reports. The client team specialized in operations management and were therefore incredibly cognizant of what data was relevant. The team then build the ingestion layer.
A major part of the engagement was referencing the best practices laid out by Microsoft to develop a workable, scalable and future-proof architecture as other data sources are onboarded. We ultimately decided on a single-workspace architecture with pseudo environments organized within a folder structure. The size of the team did not introduce security concerns between environments that would typically predate a multi-environment approach. Orchestrators were used for global processes and for each layer of the data lake with strict dependencies enforced at each level.
The app to be migrated was produced by a firm long ago and therefore required reverse-engineering with myself and a junior consultant.
Results
- Functional environment to ingest a subset of maintenance data
- Documented architectural best practices for the client and Axis
Lessons Learned
- Never assume that your compute will work the first time