چكيده به لاتين
Today, with the development of the new generation of communication networks and the advancement of mobile devices, we are witnessing a huge transformation in the emergence of new applications and applications. Limitations in the size, computing power and energy of mobile devices are among the basic challenges of this platform, which makes the devices unable to run programs that require a lot of processing resources and low latency. Mobile edge computing is a promising architecture and technology to reduce latency, improve energy consumption and efficiency of mobile devices. This architecture improves the capacity of mobile devices to perform applications by providing processing and storage resources at the edge of the network and close to the user. In this technology, computational offloading technique is used to use computing capabilities and solve the challenges of mobile devices. The process of transferring computing load from mobile devices to the cloud, edge or any computing resource near the device is called offloading, which aims to reduce energy consumption and application execution time. Computational offloading methods, in order to make offloading decisions, must consider network conditions and application features, including device mobility, which is an influential challenge in decision-making during offloading time. Due to the mobility of users, the access point around the mobile phone changes dynamically, which makes the optimality of the offloading decision difficult because the mobility can cause connection interruptions and change the access point around the mobile phone, which leads to the change of network-related parameters such as bandwidth, delay and rate The network is loaded. Therefore, awareness of mobility in the offloading process in order to make the best decision and gain the most benefit in terms of reducing energy consumption and delay will have a significant impact on the efficiency of computational offloading.
In this research, in order to reduce offloading and prevent multiple migration of user components, a mobility-aware offloading method with dual connection capability is presented. In this method, each user has a program with several components, and data is called among the components of each user. By assigning user components to two servers at the same time, we will see a reduction in latency, network overhead and possible migrations. Also, to realistically consider the mobility of users, a machine learning-based method is used to predict the mobility of users and the characteristics of user components during time slots. The objective function defined in this research is to minimize the overall computational offloading time of all users in all time slots. The total offloading latency considered in this work includes offloading time, processing time and migration time. According to the results obtained from the evaluations, it can be seen that the proposed approaches, in line with the goal of the problem, i.e. the overall reduction of offloading time compared to the compared method, have improved by an average of 28.5% in the best approach. This reduction in time has been due to the dual connection and simultaneous offloading along with fine-grained in order to reduce the cost of offloading and migration.
