A Model-Driven Approach for Software Architecture Evaluation in Large-Scale Distributed Systems

Abstract

Large-scale distributed systems (LSDS) are foundational to modern computing infrastructures, powering domains such as cloud computing, high-performance computing, and global-scale web services. However, the intrinsic complexity, dynamic scalability requirements, and technological heterogeneity of LSDS pose significant challenges for software architecture design and evaluation. Poor architectural decisions can lead to inefficiencies, scalability bottlenecks, and maintainability issues that become increasingly costly over time. This paper presents a model-driven approach to software architecture evaluation tailored specifically to the context of LSDS. The proposed methodology provides a formalized, systematic framework for assessing architectural configurations with respect to key quality attributes, including performance, scalability, and maintainability. By employing architecture description languages (ADLs) and simulation-based modeling tools, the approach enables early-stage exploration and validation of design alternatives before implementation. It incorporates both structural and behavioral modeling, allowing for the quantification of system properties under varying workloads, deployment topologies, and fault conditions.