Multi-access Edge Computing (MEC) enables computation and energy-constrained devices to offload and execute their tasks on powerful servers. Due to the scarce nature of the spectral and computation resources, it is important to jointly consider i) contention-based communications for task offloading and ii) parallel computing and occupation of failure-prone MEC processing resources (virtual machines). The feasibility of task offloading and successful task execution with virtually no failures during the operation time needs to be investigated collectively from a combined point of view. To this end, this letter proposes a novel spatiotemporal framework that utilizes stochastic geometry and continuous time Markov chains to jointly characterize the communication and computation performance of dependable MEC-enabled wireless systems. Based on the designed framework, we evaluate the influence of various system parameters on different dependability metrics such as (i) computation resources availability, (ii) task execution retainability, and (iii) task execution capacity. Our findings showcase that there exists an optimal number of virtual machines for parallel computing at the MEC server to maximize the task execution capacity.
翻译:由于光谱和计算资源的稀缺性,必须共同考虑(一) 任务卸载的基于争议的通信,以及(二) 并行计算和占用易失灵的MEC处理资源(虚拟机器),需要从综合角度对任务卸载和成功执行任务的可行性进行集体调查,操作期间几乎没有故障。为此,本信提议建立一个新的空间时空框架,利用随机几何和持续时间马尔科夫链,共同确定可依赖的MEC型无线系统的通信和计算性能。根据设计框架,我们评估各种系统参数对不同可依赖性指标的影响,如(一) 计算资源可用性,(二) 任务执行可保留性,(三) 任务执行能力。我们的调查结果显示,在MEC服务器上存在最佳数量的平行计算虚拟机器,以最大限度地实现任务执行能力。