چكيده لاتين
With the rapid growth of video applications in mobile networks and the increasing demand for high-quality, low-latency services, particularly along the path toward sixth-generation (6G) networks, the design of mechanisms to minimize content delivery delay and improve Quality of Experience (QoE) has become a key priority in content distribution architectures. Multi-access Edge Computing (MEC) has emerged as a crucial paradigm that brings computational and storage resources closer to users. In this context, cooperative caching among MEC servers has been proposed as an effective approach to enhance cache-hit ratios and reduce latency. However, the efficient implementation of cooperative caching requires an optimized communication and coordination structure among servers.
In recent years, several studies have introduced static clustering schemes for grouping MEC servers to facilitate cooperative caching. In these approaches, servers are statically organized into clusters, and intra-cluster interactions occur through interfaces such as Xn links. Although static clustering simplifies management, it fails to adapt to traffic fluctuations and suffers from performance degradation under sudden load changes or variations in user distribution. In particular, fixed cluster sizes may lead to increased latency or unnecessary inter-server overhead during content exchange.
To address these challenges, this research proposes an analytical framework and an adaptive algorithm for dynamic clustering among MEC servers. The proposed method is developed based on a mathematical analysis of delivery delay, cache-hit probability, user request rate, cluster size, Xn interface capacity, and video file size. Using queuing theory and analytical modeling of interdependent parameters, it is demonstrated that cluster size plays a crucial role in overall system performance. To solve this optimization problem, the DyMECC algorithm (Dynamic MEC Clustering) is introduced, which adaptively determines the optimal cluster size according to real-time network conditions and key system parameters; minimizing latency while avoiding excessive coordination and exchange overhead.
Simulation results show that the proposed DyMECC algorithm can reduce video delivery latency by an average of 15% under light-load conditions and up to fivefold under heavy-load conditions, compared to static clustering methods. Moreover, by controlling the inter-server traffic volume via the Xn interface, the coordination overhead is reduced by up to 45%. Sensitivity analysis further confirms that the algorithm maintains stable performance across a wide range of parameters such as request rate and file size. An additional benefit of this optimization is the stabilization of intra-cluster data exchange flows, which prevents unwanted congestion on Xn paths and contributes to smoother system operation under dynamic network loads.
