Date: February 19, 2000 (Monday)
Time: 11:30am (refreshment starts at 11:15am)
Place: 202 ECEC, NJIT
About the Speaker
Hisashi Kobayashi is the Sherman Fairchild University Professor at Princeton
Univesity, and served as the Dean of Engineering in 1986-91. Prior to
joining the Princeton Faculty in 1986, he spent 19 years with the IBM
Research Division at Yorktown Heights, NY, where he was a technical manager
in the Applied Research Department and then in the Computer Science
Department. In 1982-86, he was appointed the Founding Director of the IBM
Tokyo Research Lab., where he directed numerous research projects in
computer sciences and engineering.
He is the inventor of a high density digital recording scheme, now widely
known as PRML (partial-response coding and maximum-likelihood decoding). He
is an IEEE Fellow since 1977, the recipient of Humboldt Prize from Germany
(1979), a Silver Core Award from IFIPS (1980), and Outstanding Contribution
Awards from IBM (1971,1973). He was elected a Member of the Engineering
Academy of Japan in 1992.
He received his BE and ME degress from the University of Tokyo in 1961 and
63, and his MA and Ph.D. degrees from Princeton, in 1966 and 67. He worked
for Toshiba Co., Kawasaki, Japan as a radar system designer in 1963-65. His
current areas of research include transceiver design for wireless
communications, communication networks design and analysis, and
queuing/traffic theory.
About the Talk
The system for wireless Internet service provisioning consists of mobile
users, push/pull proxy gateways and origin Web servers. The procedure that
a mobile user browses or searches his/her desired content is rather
complicated. We may have several ways to reduce the total access time: The
proxy gateway can cache/ prefetch certain content into its storage to
satisfy mobile users locally in order to reduce the response delay; the
wireless Internet web server can provide dynamic pages with most relevant
content and links to a mobile user to reduce the number of
browsing/searching steps with the constraint of page size (i.e., limiting
the page reading time and page transmitting time in each step).
We have investigated how the "mobility tracking model" (based on a
semi-Markov process representation of mobile user behavior) and the "request
traffic model" (treated as an "emission" from the underlying hidden
semi-Markov model) that we recently developed can be applied to prefetch web
content for each mobile user for efficient wireless Internet access.
We propose a new prefetch scheme that combines access probability, response
time and updating cycle to determine the lowest average latency or the
highest hit probability. Since the required parameters in our prefetch
scheme can be simply derived from the log data of the cache, our scheme can
be implemented in practice.
For more information contact Nirwan Ansari (973) 596-3670, or check http://www-ec.njit.edu/~ieeenj for latest update.