Title: "PRESTO: A Predictive Storage Architecture for Sensor Networks"

Speaker: Deepak Ganesan, University of Massachusetts Amherst.

Abstract:

While the evolution of sensor platforms such as Motes has tracked
technology trends in computation and communication, the storage
subsystem across these platforms has undergone little change. The
comparable energy costs of storage and communication on these
platforms has raised questions about the rationale for using in-
network storage-based data management techniques for sensor networks.
I will address this question by presenting a comprehensive evaluation
of energy consumption of available flash-based storage options for
sensor networks, and demonstrate that NAND flashes are two orders of
magnitude more energy-efficient than flash memories on Motes as well
as state of the art 802.15.4 radios.

The ability to equip nodes with high-capacity, ultra-low power
storage has significant implications on a data management
infrastructure for sensor networks. I will present PRESTO, a novel
two-tier sensor data management architecture for emerging large-
scale, hierarchical sensor networks. In contrast to existing
techniques, PRESTO is a proxy-centric architecture, where tethered
proxies balance the need for interactive querying from users with the
energy optimization needs of the remote sensors. PRESTO exploits
storage trends to emphasize archival at remote sensors and
intelligent caching at proxies. Energy-efficiency in PRESTO is
achieved by extensive use of predictive techniques at the proxy to
conserve energy at the sensor nodes while simultaneously ensuring
that anomalous data events are not missed. In order to efficiently
support archival queries over distributed sensor networks, PRESTO
exposes a unified, easy to use data abstraction across numerous
proxies and remote sensors. I will present results that demonstrate
the significant benefits of PRESTO, in particular, by using ARIMA
models for predictive data management, and Interval Skip Graphs for
providing a unified data abstraction.

The PRESTO project is joint work with Prof. Prashant Shenoy and
graduate students Peter Desnoyers, Ming Li and Gaurav Mathur.

Bio:

Deepak Ganesan is an Assistant Professor at the University of
Massachusetts Amherst. He received his B.Tech in Computer Science
from the Indian Institute of Technology, Madras, India, in 1998, MS
from the University of Southern California in 2000 and his Ph.D. from
the University of California, Los Angeles,  2004. His research
interests range across a broad spectrum of topics in Wireless Sensor
Networks; his current projects include PRESTO (data management) and
SensEye (multi-tier, multi-modal camera sensor networks). He has been
actively involved in the sensor network community for the past six
years, as a member of the Center for Embedded Networked Sensing
(CENS) at UCLA, and as an intern at Intel Research, Berkeley. He is
on a number of program committees including ACM Sensys 2006, IPSN
2006 and IEEE Infocom 2006, and is an editor for the ACM Sigmobile
Mobile Computing and Communications Review (MC2R). He is a member of
the IEEE and the ACM.