“Data represents all the value that exists in the enterprise,” Richard Tarling, Managing Director & Technology Fellow, Core Engineering, Goldman Sachs, told the audience of Unqork Create 2022.
To realize that value, organizations face two yawning chasms. First, data is typically siloed and cannot move seamlessly from one system to another. Second, data modelers are too far removed from the process of designing the software that runs the models they create.
As Tarling points out, an enterprise’s data is typically spread across tens of thousands, even hundreds of thousands, of databases, with different ways of organizing the same information.
“As data flows between those databases, it is being skewed. It is being corrupted. Information is being lost,” said Tarling.
Similarly, the ideas of data modelers are often skewed as they are translated into software. Steve Hoberman, Data Modeling Author and Lecturer at Columbia University, compared it to a game of Telephone.
“Let’s say I whisper a requirement in your ear. By the time it goes through the business owners and the business sponsors and the business analysts and the data analysts and the data modelers and the data architects and other technologists, the solution sounds nothing like that requirement I whispered in your ear,” said Hoberman.
“To make it work, modeling needs to drive the process,” declared Hoberman.
From Codeless Development to Stateless Data
To tackle the problem of data, Tarling believes organizations should decouple data from the applications that they build. Instead, he said, they should deliver data as a platform in its own right. In this way, the functions that apply to the data are themselves stateless functions where the data remains at rest. There would be no need to move it, and no need to duplicate it.
“Just as we have serverless leading to codeless, the application of those technologies to data will help us solve these enterprise data challenges,” said Tarling.