Data Processing on Modern Hardware

We are exploring the architectural and design implications of the rapid evolution of hardware and cloud deployments on a wide range of systems, from databases to distributed data processing. The research’s two main directions include building systems for hybrid architectures and designing systems for the cloud. The former explores the possibilities offered by smart storage, smart NICs, CPUs, GPUs, as well as FPGAs and how to take advantage of them in real systems. The latter looks at the problems posed by cloud deployments in terms of efficiency, scalability, and generality, both in virtualized and serverless environments. In all these efforts, we strive to build innovative systems focusing on medium and long-term research problems rather than on the immediate limitations of today’s infrastructure. A big part of our research aims at creating an open source ecosystem for hardware acceleration and hybrid architectures to facilitate research and make it easier to explore the opportunities that new hardware has to offer.