thad_omura

Data-driven architectures in data center infrastructure where compute elements are moved close to the data saving massive amounts of data movement dramatically improving the efficiency and speed of modern applications that demand intense computing over vast amounts of data.

The concept that the variety, velocity and volume of data created is so large that new data center infrastructure architectures and paradigms are required, specially compute elements must be moved to be closer to the data for new data-driven applications to scale. Traditional server and storage solutions are build around an architecture and requires data be pulled from storage and fed into computing elements (typically CPUs, GPUs, FPGAs or other processing elements). After computation, results are likely written back to storage. compute2data implies a solution or direction is driving compute elements to be located right at the data storage, so the expensive move for large amounts of data does not need to happen. Computing can happen right at the data location reducing processing latency, saving power and speeding up data-driven applications.