Author: Livny, M.
Paper Title Page
Distributed Software Infrastructure for Scientific Applications  
  • M. Livny
    University of Wisconsin-Madison, Madison, Wisconsin, USA
  For more than two decades we have been involved in developing and implementing distributed software tools that have been adopted by a broad spectrum of commercial and scientific infrastructures. Ranging from state of the art rendering farms to distributed high throughput computing facilities for the LHC community our Condor software tools effectively and reliably manage large distributed infrastructures. These open source tools are distributed and supported by commercial entities in support of enterprise wide infrastructures and commercial applications. We believe that the design principals, the software development procedures and the software lifecycle practices we use are applicable to the Accelerator and LEPCS communities. Over the years we have learned the importance of information flow and policy decision points as well as an appreciation for the challenges of logging and error propagation in dynamic environments. We have adopted continuous integration methodologies that are supported by a dedicated build and test facility using multiple packaging tools and devoted effort toward installation and configuration tools. A key element of our software methodology is that we use the tools we develop. We use them to manage a campus wide computing infrastructure of more than 10K cores as well as to manage the more than 250 cores in our build and test facility that supports more than 20 software projects. We also make our infrastructure available for other communities. We work closely with our users and maintain ties with application developers that depend on our tools.  
slides icon Slides THBHMUKP05 [3.793 MB]