Big data. Big density. These interdependent factors preclude the extension of traditional information retrieval techniqu
Mobilizing Search of the Here and Now Jonas Michel, The University of Texas at Austin,
[email protected]
Motivation Big density
Big data
• • • •
• • • •
Environments that enable opportunistic wireless access to nearby resources, services, and applications both mobile and embedded in the environment.
User-generated Sensory Contextual Service/Application
Need for
• search mechanisms for human users in PNetS • a cohesive data model for pervasive computing applications
Search of the Here and Now
Internet search: Indexes data relative to its context[2]
Search of the here and now must be performed in the here and now.
Search Mechanisms There is a need for search mechanisms to help a human user find information he needs as he moves through densely populated rapidly changing information spaces. [3]
Gander ,
Gander search: Performs a query in the context
Tight spatiotemporal integration of user behavior and the immediate environment
a distributed search engine for PNetS
• Performs search directly within PNetS • Explicitly separates search from a priori indexing • Uses mobile ad hoc networking query protocols as spatial sampling strategies
human users
!" """"
!!!!!!!
!!!!
!!!!!!!
Mobile devices WSNs Smart Objects[1] RFID
Personalized Networked Spaces (PNetS)
!!!!
x
x
data Short-lived
Large volumes
Amount available far exceeds amount used
network Highly dynamic
Heterogeneous
Intermittent connectivity
These interdependent factors preclude the extension of traditional information retrieval techniques (e.g., indexing[2]) to PNetS and require the development of novel search mechanisms for pervasive computing environments.
A Pervasive Computing Data Model sensing
Application Digital Data
Physical
Rules
General-Purpose Data Model Key concern: Association of physical space with virtually accessible data and resources
Application Data Model
Spatiotemporal Trajectory
n o i t a v r e s ob
Facilitate service composition Reduce developer responsibilities Data exploits its own contextual dependencies
How is data created? How is data stored? How does data move? How does data die?
Phenomenon (P)
time
(typical approach)
Evaluation
Datum (D)
Spatiotemporal Trajectory [5] < , , , > Captures: • Initial relationship (P, D) • Pʼs expected dynamics • Dʼs actual dynamics
References
Simulated PNetS
Real-World Deployments API Methods
[6] Query Processor Tuplespace Message Center
Routing Table
Networking Server
[4],
a mobile interface for the Gander search engine
myGander
Performance analysis Overhead, latency, memory, power
Digital
Global view of PNetS Baseline for quality measurements
Service Browser / Advertiser
Clients Mobile Platform
Mobile Middleware components Verification Real-world measurements
[1] L. Atzori, A. Iera, and G. Morabito. The Internet of Things: A Survey. Computer Net., 54(15):2787-2805, 2010. [2] C. D. Manning, P. Raghavan, and H. Schutze. Intro. To Information Retrieval. 2009. [3] J. Michel, C. Julien, J. Payton, and G.-C. Roman. Gander: Personalizing Search of the Here and Now. In Mobile and Ubiquitous Systems, 2011. [4] J. Michel, C. Julien, J. Payton, and G.-C. Roman. myGander: A Mobile Interface and Distributed Search Engine for Pervasive Computing. In Pervasive Computing and Communications, 2012. (to appear) [5] J. Michel, C. Julien, J. Payton, and G.-C. Roman. A Spatiotemporal Model for Ephemeral Data in Pervasive Computing Environments. In Hot Topics in Pervasive Computing, 2012. (to appear) [6] A. Vargas. OMNeT++ Network Simulation Framework. http://omnetpp.org/, 2009.
Acknowledgement Many thanks to my advisor, Dr. Christine Julien, for her invaluable guidance and support and Dr. Jamie Payton and Dr. Gruia-Catalin Roman for their continued collaborative contributions.
mpc.ece.utexas.edu/gander/