Experimental Evaluation of a Meta-Computing-Based Self-Adaptive System for Dynamic Resource Optimisation

This study experimentally evaluates a Meta-Computing-based Self-Adaptive System (SAMCS) for workload-aware computational resource management under stable and dynamically changing workload conditions. The proposed system integrates runtime workload observation, Workload Complexity Index (WCI)-based analysis, adaptive decision-making, and dynamic resource reallocation within a closed-loop control process. A controlled simulation benchmark was conducted using three workload scenarios: low-complexity stable workload, high-complexity workload, and dynamic multi-phase workload. The adaptive system was compared with a static baseline using 30 paired runs per scenario, resulting in 180 experimental runs. Evaluation considered execution time, CPU utilization, workload complexity, adaptation behaviour, adaptation overhead, and computational efficiency. Results show that SAMCS introduces measurable execution-time overhead relative to the static baseline, while demonstrating workload-responsive resource adaptation and changes in computational resource utilization. The dynamic workload experiment further demonstrates that the system can detect changes in workload complexity and progressively reallocate computational capacity. The results therefore provide experimental evidence regarding both the capabilities and costs of workload-aware self-adaptation and highlight the importance of evaluating adaptation overhead together with resource-utilization and efficiency measures.