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

Dynamic workloads can cause static resource allocation strategies to become inefficient because resource requirements vary during execution. This study presents a simulation-based experimental evaluation of a Self-Adaptive Meta-Computing System (SAMCS) designed to dynamically adjust computational resources according to changing workload conditions. SAMCS employs a closed-loop adaptation mechanism that observes runtime workload and resource characteristics, computes a Workload Complexity Index (WCI), evaluates deterministic adaptation rules, and dynamically reconfigures worker resources and processing parameters. The proposed system was evaluated against a static baseline using three workload scenarios representing low-complexity, high-complexity, and dynamically changing workloads. Thirty paired runs were conducted for each system under each scenario, resulting in 180 experimental runs, with identical workload seeds used for corresponding baseline and SAMCS executions. The results show that SAMCS introduces measurable execution-time overhead compared with the baseline, with mean execution-time differences of 1.51 s, 9.01 s, and 9.39 s for S1, S2, and S3, respectively. Despite this overhead, the adaptive mechanism demonstrated dynamic resource reconfiguration in response to workload changes and produced substantial differences in resource utilization and computational efficiency across the evaluated scenarios. Statistical analysis using paired tests confirmed significant differences for the evaluated metrics. The findings indicate that the primary benefit of the evaluated SAMCS configuration is dynamic resource adaptation rather than direct reduction of wall-clock execution time. The study also highlights the importance of explicitly accounting for adaptation overhead when evaluating self-adaptive computing systems.