برآورد محبوبيت محتواي نرم‌افزار محور هنگام استفاده از حافظه نهان و همكاري دستگاه‌به‌دستگاه براي افزايش بهره‌وري در شبكه‌هاي خوشه‌اي

Software defined estimation of content popularity when using caching an‎d device to device collaboration to increase efficiency in clustered networks

مهندسي كامپيوتر

Abstract Continuous advancements in wireless communication technology have led to the widespread use of smart devices. In addition to computers an‎d mobile phones, tablets, wearable devices, an‎d vehicles are also connected to the Internet. This diversity has resulted in an unprecedented increase in deman‎d for high-quality content, such as real-time applications an‎d video streaming, significantly increasing backhaul link traffic in cellular networks. The use of caching networks an‎d edge caching, such as user devices, has been proposed as a solution to reduce backhaul traffic. By caching content on user devices an‎d utilizing D2D communications for content retrieva‎l, congestion an‎d delays in content delivery can be prevented. However, challenges such as limited cache capacity necessitate careful content selec‎tion for caching. Only a small portion of user-requested content is popular, yet this small portion accounts for a large fraction of network traffic. Nonetheless, content popularity is highly dynamic an‎d uncertain due to varying user tastes an‎d preferences. Less than 20% of users generate 80% of multimedia traffic. However, many existing approaches overlook heterogeneous user behavior, assuming all users have identical preferences. On the other han‎d, software-defined networking (SDN) is a new concept that has revolutionized the architecture an‎d structure of traditional networks. This technology is also used for efficient content caching. The SDN controller gathers content information from network nodes an‎d uses it to make decisions about content placement in the cache of network nodes. In this research, a software-defined content popularity estimation an‎d cache placement method in clustered networks with D2D communication capabilities is presented. Accordingly, a modular system is designed in a cellular network where the base station sends user content requests to the SDN controller. First, future user requests are predicted using machine learning algorithms. Then, content popularity in each cluster is calculated, an‎d finally, cache placement is performed based on the predicted popularity. Simulation results show that the proposed approach performs well under various network conditions an‎d offers the highest cache efficiency among the reviewed methods, with cache utilization close to 97%. Additionally, comparisons with previous methods indicate that the proposed approach offers a better hit rate with acceptable execution time, leading to reduced backhaul traffic. The proposed method is implemented in the Mininet-Wifi emulator, demonstrating an average CPU usage of 12.97% an‎d an average memory usage of 55.24%. Keywords: Content Popularity, Cache Placement, Software Defined Networks (SDN), Device To Device (D2D) Connection, Clustered Networks

6 فصل