The mission of the Forecast Systems Laboratory (FSL) is to anticipate the science and technology that will be needed by the nation's operational atmospheric, oceanic, and hydrologic forecasting services in the next five to ten years. The Data Systems Group (DSG) of the Information and Technology Services (ITS) department within FSL is responsible for the acquisition, processing and distribution of some 150 GBytes of data a day in support of this mission. This ever-increasing volume coupled with the diverse formats used to describe these data are the main problems facing developers within DSG as they try to reduce the cost and increase the efficiency of the services provided. For a full description of the data and system configurations within the Central Facility see Recent Advances in the FSL Central Facility Data Systems (Lipschutz, 2005). By the late 1990s the legacy systems, based on heterogeneous platforms running software developed within FSL using a functional decomposition design process, were becoming costly to maintain in both license fees for the various operating systems and, even more tellingly, software maintenance by the Data Systems Group. A decision was made at that time to invest in the effort of refactoring systems to achieve two major goals. The first was to switch to open source tools running under Linux and the second, to redesign and implement the data handling software using Object Oriented Analysis and Design (OOAD) techniques in order to reduce those software maintenance and operational costs. The remainder of this paper concentrates on the second element of this effort, the software architecture changes and resulting implementation that became the Object Data System (ODS). It will include examples to illustrate these changes and how they have improved a CF function within FSL.
This publication was presented at the following: