PlanetP

Summary            People            Publications            Funding

Summary PlanetP started out as a peer-to-peer (P2P) project concerned with content addressing, i.e., content search, ranking, and retrieval.  Our initial focus was on achieving a good relevance ranking for content search rather than extremely scalable key-based object location.  Thus, PlanetP adopted the novel approach of building an unstructured community on top of gossiping: peers gossip to help each other maintain an accurate local copy of the global membership directory as well as a compact summary of the content being shared by each peer.  While this approach limited PlanetP's ultimate scalability---we were targeting communities of several to ten thousand peers rather than millions or billions---it had the advantage of being much simpler to implement.**  PlanetP implemented an approximation of the TFxIDF vector space content ranking algorithm on top of the underlying gossiping infrastructure.  Simulation results showed that our approximation is competitive with a centralized implementation of TFxIDF, loosing very little ranking accuracy.

After completing a prototype of the PlanetP infrastructure, our interest has evolved more toward addressing the usability, dependability, and manageability of federated enterprise systems, rather than building open P2P systems.  Technology and inter-enterprise collaboration trends are pushing many enterprise computing environments to become highly heterogeneous and decentralized, making them more difficult to use and maintain.  Our theory is that P2P technologies can be leveraged to address many of the problems introduced by these trends.  While smaller than the global systems that typifies P2P computing, we believe that the scale, heterogeneity, and decentralized nature of these federated systems present significant challenges. For example, trace data shows over 1,200 unique MAC addresses connected to our departmental wireless network in a period of 2 years while an undergraduate file sharing community registered over 10,000 unique users in just one month. This data shows that federated enterprise systems will have to scale to at least thousands of users and their devices, even in a modest-sized organization such as our department. Concurrently, enterprises have more stringent requirements for the coordination and control of devices operating within their purview.

Our current work includes two directions:

  1. Leveraging PlanetP to build a federated file system called Wayfinder.  The goal of this work is to reduce the currently escalating data management “futz” by unifying the fragmented views of individual devices into a global namespace, providing device-independent name and content addressing.
  2. Exploring the application of coordination and control mechanisms such as Law Governed Interaction (LGI) to enterprise computing environments comprised of conglomerates of distributed heterogeneous software components.

** Curiously, to reduce lookup latency for DHT-based systems, several recent research efforts either maintained full membership information at each peer or replicated the content to achieve O(1) lookup.  These efforts essentially sacrifice bandwidth (for disseminating membership information or for replicating content) to reduce lookup latency, making them similar in spirit to PlanetP's gossiping approach.

Students Konstantinos Kleisouris
Kien Trung Le
John Murphy
Christopher Peery
Tuan Phan
Former Students Francisco Matias Cuenca-Acuna
Swarup Danashekar
Zhijun He
Chunling Hu
Faculty Thu D. Nguyen
Richard Martin
Publications Enforcing Enterprise-wide Policies Over Standard Client-Server Interactions.  Z. He, T. Phan, T. D. Nguyen.  In Proceedings of the 24th Symposium on Reliable Distributed Systems (SRDS), October 2005.
[PDF, BibTeX entry]

Self-Managing Federated Services. F. M. Cuenca-Acuna and T. D. Nguyen.  In Proceedings of the 23rd Symposium on Reliable Distributed Systems (SRDS), October 2004.
[PostScript, PDF, BibTeX entry]

Wayfinder: Navigating and Sharing Information in a decentralized world. C. Peery, F. M. Cuenca-Acuna, R. P. Martin, and T. D. Nguyen. In Proceedings of the 2nd International Workshop On Databases, Information Systems and Peer-to-Peer Computing (co-located with VLDB2004), August 2004.
[PostScript, PDF, BibTeX entry]

A Probabilistic Approach to Building Large Scale Federated Systems. F. M. Cuenca-Acuna. Ph.D. Thesis, Department of Computer Science, Rutgers University. April 2004.
[PostScript, PDF, BibTeX entry]

Autonomous Replication for High Availability in Unstructured P2P Systems.  F. M. Cuenca-Acuna,  R. P. Martin, and T. D. Nguyen. In Proceedings of the 22nd Symposium on Reliable Distributed Systems (SRDS), October 2003.
[PostScript, PDF, BibTeX entry]
[Extended version DCS-TR-509 (Updated July 2003), More results not shown on the paper]

PlanetP: Using Gossiping to Build Content Addressable Peer-to-Peer Information Sharing Communities.  F. M. Cuenca-Acuna, C. Peery, R. P. Martin, and T. D. Nguyen. In Proceedings of the 12th International Symposium on High Performance Distributed Computing (HPDC), June 2003. 
[PostScript, PDF, BibTeX entry]
[Original DCS-TR-487 (September 2002),  More results not shown on the paper]

Text-Based Content Search and Retrieval in ad hoc P2P Communities.  F. M. Cuenca-Acuna and T. D. Nguyen. In proceedings of The International Workshop on Peer-to-Peer Computing, May 2002. 
[PostScript, PDF, BibTeX entry]
[Original (longer) DCS-TR-483, More results not shown on the paper]

PlanetP: Using Gossiping and Random Replication to Support Reliable Peer-to-Peer Content Search and Retrieval.  F. M. Cuenca-Acuna,  R. P. Martin, and T. D. Nguyen. DCS-TR-494, Department of Computer Science, Rutgers University. July 2002. (This is a short position paper)
[PostScript, PDF, BibTeX entry]

PlanetP: Infrastructure Support for P2P Information Sharing.  F. M. Cuenca-Acuna, C. Peery, R. P. Martin, and T. D. Nguyen. DCS-TR-465, Department of Computer Science, Rutgers University. November 2001. (The gossiping algorithm described in this paper is out-of-date. Only read this paper if you are interested in our analysis of the "reliability" of the Information Brokerage Service.)
[PostScript, PDF, BibTeX entry]

Support/Funding PlanetP is currently supported in part by NSF grant CNS-0448070.  PlanetP was previously supported in part by NSF grants EIA-0103722 and EIA-9986046