Hardware-accelerated Networked Systems

Emerging networked systems, e.g., distributed storage and machine learning platforms, demand high-performance networking. Advanced networking hardware, including optical switches, programmable switches, RDMA NICs and smart NICs can answer this need. However, extracting the required step change in performance and functionality poses many challenges and will necessitate further hardware innovation. Hardware innovations in isolation, however, are not sufficient–they enable yet also require new architectures for the network and for applications. For example, new ultra-fast optical switching technologies will likely necessitate a step away from the traditional packet-switched network model. More flexible programs on switches will require new management frameworks. Even commodity hardware such as RDMA over Ethernet creates new problems such as congestion spreading and deadlocks. This session will bring together thought leaders in Microsoft and in academia to rethink how we co-design networked systems and applications with advanced networking hardware to fuel the cloud of the future.

Speaker Bios
Yibo Zhu
Yibo Zhu is a PhD candidate at University of California, Santa Barbara. His research interests are in datacenter networks and wireless networks. He explores new network technologies, such as 60GHz wireless, Remote Direct Memory Access (RDMA) and programmable network switches. He designs multiple novel systems that are published in top conferences and are deployed in Microsoft Azure datacenters. He receives B.S. from Tsinghua University (2011), with distinction among all college students in Beijing. He is a recipient of Microsoft Research Fellowship (2015) and UCSB Holbrook Fellowship (2011).
Senior Researcher Hitesh Ballani
Hitesh Ballani is a Senior Researcher at Microsoft Research. He designs and builds networked systems that strike a balance between clean-slate and dirty-slate solutions. His current research focuses on data center networks and rack-scale computing. His recent work on predictable data centers led to the Storage Quality of Service feature in Windows Server. He graduated with a Ph.D. from Cornell University in 2009 where he worked on network management, Internet routing, and IP anycast.
Tom Anderson
Sambhrama Mundkur
Séries:
Microsoft Research Faculty Summit
Date:
Haut-parleurs:
Yibo Zhu, Hitesh Ballani, Tom Anderson, Sambhrama Mundkur
Affiliation:
University of California, Santa Barbara; Microsoft Research

Taille: Microsoft Research Faculty Summit