CSR-PDOS: A Scalable and Self-Organizing System Architecture for Fast and Efficient Information Retrieval in Large-Scale Decentralized Networks
 

Sponsor: NSF Grant CNS-0615055, CNS-0753066

Amount funded: $225K

Principal Investigator: Dr. Duc A. Tran

Period of investigation: 9/2006-9/2009

Abstract

Project Reports (year 1, year 2, year 3)

NSF Highlight

Publications

Talks, Presentations, and Courses

Demos

Students

Collaborators

ABSTRACT

Decentralized networks offer two main advantages: no single point of failure and no capacity limitation. The more nodes added, the more potential gain in accumulative computing power and contents that the network can bring to the users. The amount of these distributed resources, however, grows tremendously over time. As such, an important problem, yet challenging, is how to manage the resources efficiently and be able to find them quickly when necessary in a network that is highly dynamic with frequent node membership changes. Many works attempt to address this problem, but only partially. If built together, they could provide a good set of services useful for the users. Unfortunately, such a combination is hardly feasible as most of these techniques differ fundamentally in system architectural design. This project develops a unified system infrastructure, that is scalable, self-organizing, and facilitates a comprehensive set of fast, accurate, and efficient information retrieval services for the users of decentralized networks. This infrastructure contains five components: (1) communication, (2) indexing, (3) search, (4) ranking/aggregation, and (5) security. The expected outcome includes a software implementation of the infrastructure with built-in technologies for these components. The development results will be disseminated via the Web, seminars, publications, and industrial collaborations. The educational plan is designed to significantly advance students' knowledge, at both undergraduate and graduate levels, and prepare them well for future professional endeavors. The project is applicable to any large-scale distributed information-sharing network deployed in the scientific, commercial, and homeland-security fields, to name some.

PUBLICATIONS (since 9/2006)

Journal (2) & Conference (8, including a Best Paper Award)

1. S. Ghadge, T. Killingback, B. Sundaram, and Duc A. Tran. A Parsimonious Statistical Protocol for Generating Power-Law Networks. IEEE Workshop on Grid and P2P Systems and Applications (GridPeer2009 - in conjunction with IEEE ICCCN 2009), San Francisco, CA, USA, August 2-6, 2009
2. Duc A. Tran and C. Pham. PUB-2-SUB: A Publish/Subscribe Content-based Framework for Cooperative P2P Networks. In Proceedings of IFIP/TC6 Networking 2009 Conference, Aachen, Germany, May 2009
3. Duc A. Tran and C. Pham. Cost-Effective Multidimensional Publish/Subscribe Services in Sensor Networks. IEEE Workshop on Localized Communication and Topology Protocols for Ad hoc Networks (LOCAN 2008 - in conjunction with IEEE MASS 2008), Atlanta, GA, USA, September, 2008 
4. Duc A. Tran and T. Nguyen. Publish/Subscribe Service in CAN-based P2P Networks: Dimension Mismatch and the Random Projection Approach.  IEEE Int'l Conference on Computer Communications and Networks, US Virgin Island, August, 2008 (Best Paper Award)
5. Duc A. Tran and K. Nguyen. Multidimensional Information Retrieval in P2P Networks.  NSF Workshop on Next Generation Software - IEEE International Parallel and Distributed Processing Symposium – IPDPS 2008, Miami, Florida, USA, April 13-14, 2008
6. T. Nguyen, K. Kolazhi, R. Kamath, S. Cheung, and Duc A. Tran. Efficient Multimedia Distribution in Source Constraint Networks. IEEE Transactions on Multimedia, 10(3):523-537, 2008
7. Duc A. Tran, S. Pamidimukkala, and P. Nguyen. A New Approach to Relevance-Feedback Image Retrieval based on Multiple Instance Learning. IEEE International Conference on Computer and Information Science (ICIS 2008), Porland, OR, USA, May 14-16, 2008
8. Duc A. Tran and K. Vu. Dimensionality Reduction with Reconstructed Boundedness in Hyperbolic Data Spaces. IEEE International Conference on Computer and Information Science (ICIS 2008), Porland, OR, USA, May 14-16, 2008
9. Duc A. Tran and T. Nguyen. A Random Projection Approach to Subscription Covering Detection in Publish/Subscribe Networks. IEEE Conference on Collaborative Computing, New York, NY, November 12-15, 2007.
10. Duc A. Tran and T. Nguyen. Hierarchical Multidimensional Search in P2P Networks.  Journal of Computer Communications, Elsevier 2(5/6):414-427, 2007

TALKS, PRESENTATIONS, AND COURSES

• Presentation slides on PUB-2-SUB (at NETWORKING 2009, also an invited talk at UMass Lowell 2/6/2009)
• Presentation slides on EZSearch
• Presentation slides on Random Projection Approach to Subscription Covering Detection in Pub/Sub Systems (talks at CollaborateComm 2007, UMass Boston seminar 2007)
  
• Course on Distributed Information Retrieval  (University of Dayton, Winter 2007)

DEMOS (of Software Systems Developed in the Project)

• LNFA: Lognormal Fitness Attachment Protocol
• iSee: a web-based image retrieval system using the concepts of Machine Learning. 
• Zigzag: a P2P networking system (applied to video streaming)
• EZSearch - a P2P search system built on top of Zigzag
  

• Charlie Pham (PhD, 9/2007 - current)
• Khanh Nguyen (PhD,  9/2007 - current)
• Phuong Nguyen  (MS, Summer 2008)
• Salini Rao Pamidimukkala (MS, 9/2006-5/2007)
• Surya Ponnapalli (MS, 1/2007-5/2007)
• Stephen Igbinovia (Undergraduate, NSF/REU, Summer 2008)
  

• Dr. Linh H. Truong (2009), Senior Scientist, IBM Zurich Research Lab (co-author work on Publish/Subscribe Techniques)
• Dr. Haiying (Helen) Shen (2009), Dept. of ECE, Clemson University (co-organize the 2009 IEEE Workshop on Grid and P2P Systems and Applications)
• Dr. Junichi Suzuki (2009), Dept. of CS, UMass Boston (co-organize the 2009 IEEE Workshop on Information Retrieval in Sensor Networks)
• Dr. Thinh Nguyen (2007, 2008), School of EECS, Oregon State University (co-author work on Heterogeneous P2P Networks and Multidimensional Search in P2P Networks)
• Dr. Sen-Ching Cheung (2008), Dept. of ECE, University of Kentucky (co-author work on Heterogeneous P2P Networks)
• Dr. Khanh Vu (2008), School of EECE, University of Central Florida (co-author work on Hyperbolic Space Dimensionality Reduction)