IDEA: Intelligent Diffusion Model for Edge Cache System in Content Delivery Network
Yicong Zhao, Zhi Li, Lingshan Kong, Song Guo, Longfei Jin, Zhao Yang, Jing Yu, Zichen Wang · 2024
Content Delivery Network is an overlay network, which distributes the content from the origin servers to dis-tributed edge service nodes, enabling users to get the required content from a nearby location, therefore, improving the per-ceived user experience. In order to accelerate content transmission speed and reduce network latency and congestion, CDN load balancing technology was developed. This technology optimizes the allocation of incoming traffic across multiple servers, ensuring equitable workload distribution and mitigating risks of server overload or underutilization. Therefore it improves the respon-siveness of the website, reduces communication overhead, and maximizes the efficiency of computational resources. However, the balancing methods proposed by academia use various algorithms or approaches, but these algorithms often fail to account for the trade-off between communication costs and latency especially when the service machine is down. In recent years, most load balance techniques in the industry have primarily used edge node storage resources while neglecting computational resource utilization. To reduce both communication costs in routing and the waiting times of end-users, it is essential to direct user requests to the most optimal servers available. This study presents the implementation and design of IDEA, an Intelligent Diffusion Model for Edge Cache System in CDN, which reduces the cost of communication link and the time required to fetch distributed contents by introducing Long-Short Term Memory Network combined with a new meta-heuristic algorithm. We then compare it to other algorithms available in this paper. Experimental results show that this algorithm exceeds the performance of current methods across the key metrics outlined above.