A Distributed Computing Environment for Multidisciplinary Design

The Framework for Interdisciplinary Design Optimization (FIDO) project has the goal of developing a general distributed computing system for executing multidisciplinary computations on a networked heterogeneous cluster of workstations and vector and massively parallel computers. The concept being used for FIDO is course-grained parallelism, with instances of disciplinary codes (aerodynamics, structures, etc. for an airplane design problem) running on separate processors (including fine-grain parallel computers), under control of an executive on another processor, and exchanging data through a centralized data manager (on yet another processor). To allow the user to monitor the progress of the design iterations, the system includes a graphical user interface (which tracks the execution of codes performing the design iterations) and a separate process, called Spy, which allows a user to extract and plot data produced during current and previous design cycles, and then to steer the design process by changing appropriate control data. The software is written in modular form to ease migration to upgraded or completely new problems. In its current state of development, FIDO is being applied to a highly simplified case of a High-Speed Civil Transport design, involving a simplified problem with very few design variables. However, it has already demonstrated the ability to coordinate multidisciplinary computations and communications in a heterogeneous distributed computing system.