Research in Internet-scale Computing: The Berkeley RadLab Perspective

Randy H. Katz

Building modern scalable Internet services -- for example, the next
YouTube -- requires a massive development and deployment effort in
terms of processing, networking, and applications infrastructures and
frameworks. To build the next EBay requires building a company the
size of EBay. Is there a better way? In this talk, I present a new
research vision that seeks to apply machine learning techniques to
manage resources in complex Internet service environments. The goal
is to develop technology that can support a small team --ultimately a
single individual -- in developing, deploying, and operating
Internet-scale services that must scale rapidly. At its core, such a
research agenda encompasses the intersection of research in
applications, systems architecture, operating systems, networking,
and machine learning. This agenda is actively being pursued at the
Berkeley RadLab -- Reliable Adaptive Distributed Systems Laboratory,
a collaborative activity of six faculty and over thirty graduate
students. A particular current focus of RadLab is on intelligent
adaptive management of datacenter resources, such as power and
network quality of service within enterprise environments. I will
present recent approaches and results in these areas.

Short Bio:
Katz received the A.B. degree from Cornell, and the M.S. and Ph.D.
from Berkeley. He joined the Berkeley faculty in 1983, where he is
currently the United Microelectronics Corporation Distinguished
Professor. In May 2007, he received an honorary degree from the
University of Helsinki. (:-)) He is a Fellow of the ACM and the IEEE,
and a member of the National Academy of Engineering and the American
Academy of Arts and Sciences. He has over 250 refereed publications,
and has supervised 43 M.S. theses and 37 Ph.D. dissertations. His
recognitions include thirteeen best paper awards, three best
presentation awards, an Outstanding Alumni Award, the CRA Outstanding
Service Award, the Berkeley Distinguished Teaching Award, the Air
Force Exceptional Civilian Service Decoration, the IEEE Reynolds
Johnson Information Storage Award, the ASEE Frederic E. Terman Award,
the ACM Karl V. Karlstrom Outstanding Educator Award, and the ACM
Sigmobile Outstanding Contributor Award. With colleagues at Berkeley,
he developed RAID, a $15 billion/year storage industry sector. While
on leave in 1993-1994, he established whitehouse.gov and connected
the White House to the Internet. His current research interests are
Reliable, Adaptive Distributed Systems supported by new services
deployed inside the network.