The world has become connected, and every year Zetabytes of data are transferred from users to the cloud. This trend, however, cannot sustain. Already the amount of data generated by users' applications exceeds the processing capacity of any single computer. An emerging research area, In-network computing, tries to attend to these limitations. In-network computing refers to the offloading of standard applications to network devices, where processing is done as the data traverses the network. This talk will introduce a set of in-network computing applications and frameworks that we have developed, ranging from network services (e.g., DNS) to consensus (e.g., Paxos) and machine learning. I will focus on the challenges in developing and using such applications through practical examples. Still, in-network computing offers immense performance and power saving benefits, and the talk will conclude with a discussion of future research directions.
Please email for a
Noa Zilberman is an Associate Professor in Engineering Science, leading the Computing Infrastructure research group (https://eng.ox.ac.uk/computing/
) at the University of Oxford. Her research expertise is in programmable devices and networked-systems. Previously, she was a Fellow and an Affiliated Lecturer at the University of Cambridge. Before moving to academia, she spent close to 15 years in industry, last as a Hardware Manager and Chip Architect at Broadcom.