Computing is bottlenecked by data. Modern computing systems and data centers have large difficulties in dealing with data, taking advantage of the vast amounts of data,and exploiting different properties of application data. Our recent work with Google shows that more than 60% of entire system energy is consumed on moving data in the memory hierarchy.
Handling data well requires designing system architectures based on three key principles: data-centric, data-driven & data-aware. Let us discuss recent research that aims to fundamentally reduce memory latency and energy, and practically enable computation close to data. We will discuss how to enable adoption of such fundamentally more intelligent architectures.
