Data-centric computing

Content sourced from Wikipedia, licensed under CC BY-SA 3.0.

Data-centric computing is a new way to design information systems. It treats data as the main asset and stores it independently from the programs that use it. This lets applications be upgraded or replaced without moving data, saving time and reducing risk.

The change comes as data grows, especially unstructured data like text, images, and videos. About 90% of new data is unstructured. A 2018 finding shows many organizations manage billions of files across many servers. Businesses want faster insights with big data analytics and machine learning, but old systems can’t scale easily to petabytes and beyond.

Data-centric computing shifts the focus from apps to data. It combines new hardware and software so data movement and data analysis matter more than keeping a fixed set of applications fast. Data becomes the permanent value; applications come and go as needed.

Hardware must scale massively, move data quickly, and handle large calculations. Software is redesigned so programs are short-lived and updated often. Instead of working on small data subsets, software analyzes all available data, and microservices access data to perform calculations and share results quickly.

Researchers note that data quality affects model performance, so data-centric approaches aim to improve results by improving the data itself, not just the algorithms.