DISCO: A Framework for Classification and Selection of Software Connectors for Highly Distributed and Voluminous Data-intensive Systems

Student: Chris A. Mattmann, USC

Collaborators: David Woollard, USC; Reza Mahjourian, University of Florida; Trevor Johns, USC

Advisor: Nenad Medvidovic, USC/ISR

Abstract: Selecting software connectors for large-scale data distribution systems is more of an art form than a principled methodology, and frequently involves consulting organizational gurus, who base their selections on "hunches", and past experience with similar systems built. Such connector selections are becoming increasingly important as data volumes continue growing, and as disk space becomes cheaper. Large, previously closed loop data archives, are now challenged with the task of distributing their large amounts of data to users around the world, separated by geographic heterogeneity, network latency, in a manner that is performant, and that satisfies user-imposed constraints on the data distribution activities. With the abundance of available software connector technologies that deal with data movement (bbFTP, GridFTP, UFTP, HTTP/REST, SOAP, RMI, etc.), what connector (or set of connectors) is best suited for a particular data distribution scenario?

We present a software architecture-based framework for automatic connector classification, selection and analysis that answers the above challenges accurately, expressively, and efficiently, allowing an architect to explore the trade-off space when architecting large data distribution systems. Our framework, dubbed DISCO, is evaluated in the context of 30 real world data distribution scenarios, spanning the domains of planetary science, cancer research and earth science. Our evaluation of DISCO has validated our early hypotheses that the development of such a framework was possible, and ultimately useful to users and builders of data distribution systems.

Bio:

Chris Mattmann is a Ph.D. candidate in the Computer Science Department at the University of Southern California, where he is a member of the Software Architecture Research group, advised by Dr. Nenad Medvidovic. His dissertation research has investigated software connectors and their properties in highly distributed and voluminous data-intensive systems. Chris is scheduled to defend his dissertation in July 2007. In parallel to his research at USC, Chris works full time at the Jet Propulsion Laboratory, as a Key Staff member in the Modeling and Data Management Systems section, where he is managed by Daniel J. Crichton.