Intelligent Big Data Lake Governance

When you have data, and data which is flowing fast with variety into the ecosystem, the biggest challenge is governing that data. In traditional data warehouses, where data is strucured and the structure is always known, creating processes, methods, and frameworks is quite easy. But in a big data environment, where data flows fast while inferring run time schema, the need to govern data is at run time.

When I was working with my team to develop an ingestion pipeline and collecting ideas from the team and other stakeholders on how the ingestion pipeline should be, one idea was common: can we build a system where we can analyze what changed overnight in a feed structure. The second requirement was finding the pattern of the data, e.g. how could we find out that a data element was a SSN numer, a first name, etc., so that we can tag the sensitive information at run time?

The core data governance team was struglling with how to maintain information about each and every element from all the feeds, as in a data lake there are more than 3,000 feeds. And information about each and every elemnt in those 3,000 feeds was equivalent to maintaing the information of more than 100,000 elements. That's a quite a lot of work; is there any way to do it? I had a thought that if I needed to build an ingestion frameowerk for an ecosystem like Hadoop, where updates are challenges, data governanace methods and rules should be applied during ingestion only. This problem needed to be solved for structured and unstratucred feeds. In any organization, the bigger part of data is structured.

For structured data, we have built a metadata repository-based ingestion frameowrk. Before ingestion,  MySQL-based feeds of metadata, elements' metadata, user roles and access metadata, and SLA metadata are used to fill in the metadata repositry. For each and every element in the metadata repository, data governance rules and methods (PII, master data, critical data element, pattern matching, business rules) are defined.

Once all the metadata is defined, on the arrival of the feed, it gets ingested as per the defined rules. During ingestion, all the rows were identified with valid and invalid elements. And for each row a trust factor was created, which was the indicator of how much the row can be trusted on a scale from 1 to 100. If it's 70, that means 70% of the elements in the row could satisfy the data governanace data quality rules.