Research Interests

 My general areas of interest are sensor networks, distributed systems and data dissemination.

Following is a brief description of my work with links to relevant publications.

Wireless and Sensor Networks

 Recent technology trends that have resulted in a broad spectrum of sensors, wireless radio technologies, and embedded sensor platforms with varying capabilities. The advancements can be exploited to design sensor network applications that are hierarchical with multiple tiers, where each tier employs sensors with different characteristics. Multi-tier networks are not only scalable, they offer a number of advantages over simpler, single-tier unimodal networks: lower cost, better coverage, higher functionality, and better reliability. The design of such mixed networks raises a number of new challenges that are not adequately addressed by current research. In my thesis, I address issues related to the design and implementation of a multi-tier camera sensor network, study the benefits of location prediction techniques and develop a automatic camera calibration technique using position sensors.
    Approximate Initialization of Camera Sensor Networks [pdf]
    Purushottam Kulkarni, Deepak Ganesan and Prashant Shenoy
    Proceedings of 4th European Conference on Wireless Sensor Networks (EWSN 2007).
    Snapshot: A Self-Calibration Protocol for Camera Sensor Networks [pdf]
    Xiaotao Liu, Purushottam Kulkarni, Prashant Shenoy and Deepak Ganesan
    Proceedings of IEEE/CreateNet BASENETS 2006.
    SensEye: A Multi-tier Camera Sensor Network [pdf]
    Purushottam Kulkarni, Deepak Ganesan, Prashant Shenoy and Qifeng Lu
    Proceedings of the ACM Multimedia 2005.
    Nominated for the best student paper award..
    The Case for Multi-tier Camera Sensor Networks [pdf]
    Purushottam Kulkarni, Deepak Ganesan and Prashant Shenoy
    Proceedings of NOSSDAV 2005.
    Exploiting Overlap for Provisioning of Access Points in Wireless Networks [pdf]
    Purushottam Kulkarni and Prashant Shenoy
    Technical Report, October 2004, University of Massachusetts.

Web Caching

 An important aspect of web-based data dissemination is caching. The vital tradeoff on caching based systems is consistency vs. bandwidth requirements. We have studied the use of Leases to maintain delta-consistency at proxy caches. The architecture used a hierarchy of cooperative proxy caches and studied the performance and scalability of the approach. The idea of leases was also extended to study overheads and benefits when clients move locations.
Another related problem we looked at was that of simulating large scale CDNs. Simulations of CDNs requires maintenance of state for each simulated proxy. With large scale simulation the memory requirement becomes and bottleneck. We developed approximate state maintenance techniques using Bloomfilters and study their benefits of reduced memory usage and impact on accuracy.
    Scalable Techniques for Memory-efficient CDN Simulations [pdf]
    Purushottam Kulkarni, Weibo Gong and Prashant Shenoy
    Proceedings of the 12th WWW Conference (WWW 2003).
    Handling Client Mobility and Intermittent Connectivity in Mobile Web Accesses [pdf]
    Purushottam Kulkarni, Prashant Shenoy and Krithi Ramamritham
    Proceedings of the Conference on Mobile Data Management (MDM 2003).
    Scalable Consistency Maintenance in Content Distribution Networks using Cooperative Leases [pdf]
    Anoop Ninan, Purushottam Kulkarni, Prashant Shenoy, Krithi Ramamritham and Renu Tewari
    IEEE Transactions of Knowledge and Data Engineering, July/August 2003.
    Cooperative Leases: Scalable Consistency Maintenance in Content Distribution Networks [ps]
    Anoop Ninan, Purushottam Kulkarni, Prashant Shenoy, Krithi Ramamritham and Renu Tewari
    Proceedings of the 11th WWW Conference (WWW 2002).
    Chosen as the best paper in the performance category.
    Implications of Proxy Caching for Provisioning Servers and Networks (extended version) [pdf]
    Mohammad S. Raunak, Prashant Shenoy, Pawan Goyal, Krithi Ramamritham and Purushottam Kulkarni
    IEEE Journal on Selected Areas in Communications (JSAC), Sept. 2002.

Other Work

 Data storage and communication capacities are ever-increasing, but significant benefits can still be obtained by reducing the number of bytes required to store or transmit an object. At IBM Research Labs, as an intern, I worked on developing a technique called REBL, Redundancy Elimination at the Block Level. The technique combines delta-encoding and duplicate block suppression for efficient and effective elimination of data redundancy.

QoS requirements for multimedia applications can be supported using both OS-level and Middleware techniques. I briefly worked on studying the benefits and tradeoffs of each approach.

Decision tree learning, is an extensively used supervised learning algorithm. As part of my Master's thesis, i developed a technique to parallelize the algorithm. The technique primarily used the MPI library for message passing and coordination.
    Redundancy Elimination within Large Collections of Files [pdf]
    Purushottam Kulkarni, Fred Douglis, Jason LaVoie and John M. Tracey Proceedings of the USENIX 2004 Annual Technical Conference, June 2004.

    Middleware versus Native OS Support: Architectural Considerations for Supporting Multimedia Applications [pdf]
    Prashant Shenoy, Saif Hasan, Purushottam Kulkarni and Krithi Ramamritham
    Proceedings of IEEE Real-time Technology and Applications Symposium (RTAS'02), September 2002.

    Some methods for parallelizing decision tree learning [pdf]
    Kulkarni, Purushottam (May 2000).
    Master's Thesis, University of Minnesota Duluth, Computer Science Department.