Abstract

The world has become connected. Every year Zetabytes (10^21) of data are transferred through networks, and processed by increasingly powerful computers. This trend, however, can not sustain. Already the amount of data generated by user applications exceeds the processing capacity of any single computer. In recent years a new class of network devices, programmable network devices, has been introduced. The programmability of these devices allows us today not only to flexibly support more functionality within the network, but also to move ``standard’’ applications to run within the network. This talk will focus on this type of processing within the network: In-Network Computing. Starting from an introduction to programmable network device, I will explore the rise of In-Network Computing, the need for it, and the challenges that are yet to be faced. With applications running within the network achieving orders-of-magnitude performance improvement, without incurring additional hardware or power cost, I will discuss the applicability of In-Network Computing to a range of applications and research disciplines.

Bio: Noa is a Leverhulme Early Career Fellow at the Computer Laboratory, where she works on high performance networed-systems. Previously, Noa had spent 15 years in industry, last as a Senior Principal Chip Architect at Broadcom.