Booth, Thomas M., authorGhosh, Sudipto, advisorHerber, Daniel, committee memberBlanchard, Nathaniel, committee memberVijayasarathy, Leo, committee member2025-09-012025-09-012025https://hdl.handle.net/10217/241905https://doi.org/10.25675/3.02225Data systems consist of a network of communication channels, software that processes, creates, or transmits data across these channels, and the hardware that runs these applications and generates data. Software-centric data systems rely on software to define system behavior, security, and other capabilities. To keep up with rapidly shifting system requirements, many successful organizations utilize software Development and Operations (DevOps) principles to increase software quality and throughput of new software capabilities. Many organizations have modern data systems that include cloud resources and 49% of these organizations lack a Financial Operations (FinOps) team for cloud management and optimization. A third of these organizations spend more than $12 million/year on cloud costs and more than half are regularly 18% over their annual cloud budget. Similar resource management issues exist for aircraft avionics data systems which also cost organizations millions of dollars a year. Adopting a practical system analysis and optimization methodology is one solution to these organizational budget issues. Model-Based Systems Engineering (MBSE) enables the design, analysis, and optimization of complex systems and has been in practice since the late 20th century. The Systems Modeling Language (SysML) is currently the most common graphical specification language used with MBSE methodologies. Unfortunately, there is insufficient research into MBSE and SysML that show a positive Return On Investment (ROI) for the design, development, sustainment, and optimization of software-centric data systems. Additionally, to keep up with the rapid deployment of software capabilities, a practical MBSE methodology and modeling approach for data systems would need to include DevOps principles. We present Yet Another MBSE Methodology (YAMM), a novel and practical MBSE methodology towards the design, development, optimization, and sustainment of software-centric data systems. YAMM has refined and extended the Harmony agile MBSE process (Harmony aMBSE) to provide a more prescriptive and tailored methodology to solve data system specific issues. We also present our novel Unified Modeling Approach (UMA) which combines graphical-based system specifications and simulations into a single, empirically-derived system model which accurately predicts data system resource usage, cost, and performance. Together, YAMM and UMA integrate MBSE with DevOps and FinOps principles to improve data system performance and reduce costs using continuous and empirically-driven feedback throughout the system acquisition and sustainment phases. We demonstrate and evaluate YAMM and UMA using case studies of two different types of data systems that include operationally relevant empirical data. The first case study implemented the YAMM framework to enable the training, selection, and analysis of a sensor fusion Machine Learning (ML) model for a legacy aircraft sensor data system. Results show the value of using compute, memory, and networking resource consumption in addition to performance and accuracy as evaluation criteria for ML model analysis and selection. YAMM and UMA are demonstrated on the second case study towards the design, development, optimization, and sustainment of a hybrid cloud analytical data system prototype in Amazon Web Service (AWS). The non-optimized prototype system had a quadratically increasing cost with 1 and 5 year cost estimates of $1.9M and $27.7M. Results showed that the unified model had a Root Mean Squared Error (RMSE) of $865 when compared against AWS cost data. The Pareto-optimal design showed a positive YAMM ROI after 361 days with an estimated 5 year savings of $12.7M without reduction in system performance. Our UMA is limited to systems without physics-based or entity-based simulation requirements. However, results show that implementing YAMM with UMA can reduce costs, increase performance, and provide insights into dominant system characteristics for the design, development, optimization, and sustainment of software-centric data systems.born digitaldoctoral dissertationsengCopyright and other restrictions may apply. User is responsible for compliance with all applicable laws. For information about copyright law, please see https://libguides.colostate.edu/copyright.machine learningsimulationhybrid cloudavionicsmodel based systems engineeringText