Space Surveillance Network and Analysis Model (SSNAM) Performance Improvements

The Space Surveillance Network and Analysis Model (SSNAM) is an Air Force Space Command (AFSPC) model, which provides the capability to analyze and architect Space Surveillance Network (SSN) Force Structure. To provide these capabilities SSNAM supports two types of simulations: Catalog Maintenance, and Special Events (Launch, On-Orbit Events, and Breakup). There are many configuration options available with SSNAM: models for all the sensors currently in the SSN to include space based and ground based sensors, hours of operation by sensor, track capacity by sensor, models for sensors yet to be created, user defined weather conditions, National Aeronautical and Space Administration catalog growth model including space debris, and solar flux just to name a few. SSNAM is a large software system. It is written in Java, C/C++, and FORTRAN (77 & 95), represents over a million lines of code, and employs a web-based, loadsharing architecture to decrease simulation runtime. Catalog Maintenance simulations are both computationally and input/output (I/O) intensive. A typical Catalog Maintenance simulation (10K to 35K satellites simulated over a 90 day period) will generate over a terabyte of data, during the course of a simulation, which is reduced down to approximately 1.5 gigabytes. Depending on simulation configuration, runtimes can range from 12 to 48 hours on a 16 node, PC network cluster. Because of the high computational demands of SSNAM Catalog Maintenance simulations and the anticipation of transitioning SSNAM to model the maintenance of an special perturbation (SP) catalog, the SSNAM system was ported to run on Maui High Performance Computing Center (MHPCC) platforms. This port resulted in at least a three-fold increase in performance for all currently parallelized processing in SSNAM. This paper provides an overview of the SSNAM application, its web based, load sharing architecture, the effort involved with porting Java and FORTRAN to MHPCC platforms, the approach and implementation for parallelizing the SP Tasker, and the resulting performance gains.