On November 5th I attended and presented at the Midwest Architecture Community Collaboration (virtual) conference. My topic was on why graph embedding will become a ubiquitous feature of knowledge graphs. But that is not what this blog is about. This post is about the topic of how Enterprise Knowledge Graphs (EKGs) can promote more adaptable enterprise architectures through taking a Systems Thinking approach to enterprise IT strategy.
The keynote speaker was George S. Paras, the Managing Director of EA Directions and Editor-in-Chief of Architecture & Governance. He is a well-regarded enterprise architect. Although I would not classify him as a graph evangelist (yet), he really knows his systems thinking. And systems thinking and EKGs is the topic of this blog.
His keynote title was: Adaptable Architecture: Building Resilience in a Time of Change. Here are just some of the quotes from his presentation. I hope they inspire you to integrate Systems Thinking and EKGs into your IT Strategies for 2021.
The Problems of Early Optimization for Cost and Query Performance
“Fragility happens when we put too much focus on the optimization of cost and speed.”
I love this quote. In a single sentence, it summarizes many of the problems we have seen in enterprise data strategy. We have seen many companies build rigid data warehouse structures around massive denormalized fact tables at the center of star and snowflake systems. These designs popularised by Kimbal and Inman were very popular in the 1990s. But we must recall they did this because these structures keep JOIN operations to a minimum. This was before distributed native labeled property graphs were available. However, if the designers can’t get teams to agree on what facts are and what the dimensions are then every team builds another data warehouse and you get massive duplication of effort. Costs go up and agility drops. Well stated George!
The solution here is to avoid storing enterprise data in a relational database unless you have a clear business reason to do it. An architect’s unwillingness to accept the failure of the relational model for enterprise analytics is not a good excuse for not adapting to change. A single enterprise knowledge graph, with good highly normalized models of the truth, will always be more resilient to change than a single model optimized for a defect in the performance of a database archtiecture.
Systems Thinking and Enterprise Strategy Alignment
Take a “whole of company” approach to guide consistency of thought, principals and approaches focused on company-wide strategic directions.
Wow! This quote is also right in line with so much that is critical in building IT strategies that work. The terms “consistency of thought” and “principals” both reflect the idea that there should be an underlying pattern of values that bring IT strategies together. And the focus on company-wide direction also speaks to the fact that the enterprise graph strategies we are building must be carefully interwoven with the company goals. EKG strategies themselves must never be islands of thought.
Building Interconnected Data Models
Build holistic cross-company inter-connected models for business, information, technology, solutions, security, risk, reliability, and agility.
This quote speaks to the fact that so many organizations are driven by Conway’s Law: the data models reflect the design of the org chart, not what is best for the customer. Every business unit has its own database and they don’t focus on a single enterprise knowledge model. Business units spend time trying to prove that their view of the world is the best and they try to steal resources from other areas to improve their own scope. They don’t really work together to build holistic models that serve the customer.
Visualization of Enterprise Resources as an Interconnected Graph
Visualize the enterprise as many interconnected parts.
George is also a big fan of visualization tools to help people think outside of their department-level concerns. It is only when they start to ask the hard questions such as “What insights can we gain if we put data from System A and System B together?” that we get our staff to understand what questions are not being answered today.
Thanks to the MACC 2020 Conference Volunteers
I also want to thank the organizers of the MACC 2020 conference, especially Youssef Haddad and Marina Kerbel. They are unpaid volunteers who are really trying to make the world a better place through great architecture. A good example of this is how the talk I gave last year on enterprise knowledge graphs inspired Erik Erickson from Hennepin County to build a knowledge graph that looked for gaps in food shelves that serve the county. Hopefully, I will find time to talk about the great work Erik is doing with graph technologies in the future.
If you are fans of good architecture in your enterprise, please consider becoming a sponsor of future MACC events.
Here is their web site: https://midwestacc.com/