Welcome to the Handle PM Project

Recent technology advances have enabled wide area applications with Internet accessible sources. An example of an emerging wide area application involves the DOI Foundation and the community of publishers. This application utilizes the Handle protocol for location independent identifiers for digital objects. DOI is a system for identifying and exchanging intellectual property in the digital environment. It is expected that this application must scale to support tens of millions of Handles and thousands of content servers, representing the digital content managed by the publishing community, and large numbers of Handle clients.

A factor that impacts the success of such applications is their ability to monitor and predict client-side performance, in a scalable manner, given disparate client and content server characteristics, e.g., dissimilar bandwidth. Our objective is to explore open technology and tools for passive performance monitoring, to complement the current commercial solutions based on proprietary technology for specific platforms and services such as Appliant and Keynote. We plan to use shared passive information gathering techniques, as was used in constructing an access cost Catalog for WebSources

Our research objective is to develop scalable algorithms to determine the placement of some (possibly optimal) number of performance monitors (PM). Each PM would be associated with a cluster of clients and content servers, and would monitor and predict end-to-end access costs for the cluster. Each PM would develop a latency profile (LP) that will provide coverage for the cluster of clients and content servers. The concept of coverage is constructed to reflect prior results on Internet distances (IDMaps), BGP database information, and points of congestion. Informally, a latency profile (LP) provides coverage for a cluster of clients and content servers provided that the Internet distances between each pair of clients and content servers, in the cluster, is comparable. In addition, the path(s) between the pairs of clients and content servers in a cluster represented by a PP must be comparable with respect to points of congestion.

More information about this project is available in PDF

Sponsors

• Corporation for National Research Initiatives
• The Information Technology Research program and the Information and Data management Program / CISE , National Science Foundation
• Smith School of Business and the Institute for Advanced Computer Studies
• Information Technology Office, Defense Advanced Research Projects Agency (DARPA)
• School for Information Sciences University of Pittsburgh.
• Faculty of Industrial Engineering & Management, Technion - Israel Institute of Technology, Haifa, Israel. .

(Go Back )

(top of the page)