Experimental Systems Research @ ND

 
 
 

NEWS: Congratulations to Dr. William Acosta for successfully defending his thesis

Research Focus

Our group is primarily interested in research issues in accessing large objects from resource constrained environments. We typically focus on multimedia objects. Our work is experimental in nature, we draw on techniques from Operating Systems and Computer Networks. Lately, we are focusing our efforts on a scalable and resilient multimedia storage for sensor like deployments called Hydra. Our system consists of a large number of video capture units and storage bricks, connected via wired and wireless networks. Our system should scale to tens of thousands of nodes, for storage capacity in the range of peta-bytes. We are focusing on the following research challenges:

  1. BulletPeer-to-peer overlays: We are developing Makalu, an unstructured p2p overlay algorithm that seeks to choose neighbors that maximize the node expansion from the local node’s neighborhood while minimizing the cost to its neighbors. Although this algorithm does not guarantee any minimum bounds for the general expansion of the overlay, in practice, it achieves connectivity properties that are similar to a known good expander graph. We use techniques from spectral graph theory to show that the structural and connectivity properties of Makalu overlay graphs are similar to those of k-regular random graphs.

  2. BulletName space management: Given the volume of information available in our target p2p system, we are working on developing ways to name all the objects and access them in an intuitive fashion. For now, we are focussing our analysis on a cooperative iTunes share like environment.

  3. BulletData management: Nodes are transient, the objects are not. We expect our p2p storage to exist for decades if no centuries. We are investigating the technologies necessary for a self-managing system that can work with components which are unreliable in order to achieve the system longevity goals.

  4. BulletSecure provenance aware storage: Our storage infrastructure cannot rely on reliable and centralized components. We are developing our storage to be completely distributed and long lived. We are investigating the security implications of this scenario. (more info)

  5. BulletPower Aware Computing This project addresses the issue of energy-aware information access from mobile clients. (more info)

  6. BulletContent adaptation using Quality Aware Transcoding The primary focus of his research is in enabling multimedia web access from slow, expensive and congested networks using application level techniques. We utilize transcoding as the enabling technology. (more info)

  7. BulletMultimedia Sensor research Many of the components discussed above will be integrated into a large scale multimedia storage that is designed to store objects for long durations. We hope to use this infrastructure to develop retrospective surveillance systems. (more info)

Research sponsors

  1. Bullet   National Science Foundation

  2. Bullet    Defense Intelligence Agency

  3. Bullet    HP

 

Last modified: Sunday, July 06, 2008