Earlier this month, we had a stimulating conversation with colleagues from another Scottish university, in which we compared our approaches to structuring our core data.  We share the goal of making our core data as valuable as possible, in a world where data analytics is of increasing importance.  

Much data analysis now manipulates data in the form of graph structures.  These graphs are not the line charts we drew at school, but a mathematical structure of nodes (data entities) and vertices (relationships).  The topic of Graph Theory is the basis of much computer science and software design and is being increasingly used in the analysis of business intelligence data. 

Compared with traditional relational databases, graph structures offer more flexibility in the face of change, better support for streams of activity data (such as clicks on a website, or readings from sensors), and better support for more complex analysis and visualisation.  Relational databases remain important, particularly for transactional systems such as the student record and finance systems, but much of business intelligence and data warehousing is moving to graph-structured data. 

Our colleagues at the other university have an established data warehouse based on relational data structures, and they’re now looking at how to add a layer that will present this data in a graph-based structure to support more data analysis techniques. In contrast, our data warehouse is taking the opposite approach. We use graph-based structures as the foundational model and have added a layer above that to present some data back in the relational structures that are more familiar to our data analysts.   

These different approaches reflect the different stages of maturity at our two institutions.  The other institution’s data warehouse has been running for fifteen years, includes all their core data down to the level of individual financial transactions, and was built on the best technology of the time.  Our warehouse is newer and was able to use newer approaches from the start, but still has to support our existing investment in older business intelligence tools and techniques. Both institutions are adopting approaches that will maximise flexibility in the years ahead.  

You can read more about the difference between graph data and relational data in various articles on the web, such as the one here:  https://searchdatamanagement.techtarget.com/feature/Graph-database-vs-relational-database-Key-differences

Future-proofing our core data / Enterprise Architecture by blogadmin is licensed under a Creative Commons Attribution CC BY 3.0