Program Placement Optimization for Storage-constrained Mobile Edge Computing Systems: A Multi-armed Bandit Approach
Program Placement Optimization for Storage-constrained Mobile Edge Computing Systems: A Multi-armed Bandit Approach
复制标题
DOI：
10.1109/wowmom51794.2021.00028
复制
发表时间：
2021-06
期刊：
影响因子：
--
通讯作者：
Ming Feng;M. Krunz
中科院分区：
文献类型：
--
作者：
Ming Feng;M. Krunz
移动的边缘计算（MEC）是支持具有严格延迟要求的计算密集型移动的应用的有前途的技术。随着MEC应用变得更加多样化和复杂，对于具有有限存储的边缘节点（EN）来说，保持所有任务的程序代码变得更具挑战性。在本文中，我们研究了存储有限的MEC系统中的程序放置和用户关联问题。制定的问题作为一个顺序的决策问题，我们首先推导出一个单一的EN的解决方案，通过将制定成一个多臂强盗（MBA）的问题，并解决它通过汤普森采样（TS）算法。然后，我们提出了一个解决方案框架的多EN的情况下，我们将原来的问题分解成三个子问题，并解决它们与低复杂性的方法。第一个子问题是学习任务的流行度，我们也制定了一个MAB问题，并通过TS算法解决它。第二个子问题是给定用户关联下的程序布局优化问题，我们提出了一个贪婪算法来解决这个问题;第三个子问题是用户关联问题，我们提出了一个基于对偶分解的方法来解决这个问题。仿真结果表明，我们所提出的方案实现的平均延迟是30%至100%，低于两个基准计划，平均不到10%，高于下限。
Mobile edge computing (MEC) is a promising technology to support computationally intensive mobile applications with stringent delay requirements. As MEC applications become much more diverse and complex, it becomes more challenging for an edge node (EN) with limited storage to keep the program codes of all tasks. In this paper, we investigate the problem of program placement and user association in storage-limited MEC systems. Formulating the problem as a sequential decision-making problem, we first derive the solution for a single EN by transforming the formulation into a multi-armed bandit (MBA) problem and solving it via a Thompson sampling (TS) algorithm. We then propose a solution framework for the multi-EN scenario, where we decompose the original problem into three subproblems and solve them with low-complexity approaches. The first subproblem is to learn the task popularity, which we also formulate as a MAB problem and solve it via a TS algorithm. The second subproblem is optimizing program placement under a given user association and we propose a greedy algorithm to solve it. The last subproblem relates to user association, which is solved by a dual decomposition-based approach. Simulation results show that the average latency achieved by our proposed schemes is 30% to 100% lower than two benchmark schemes and is on average less than 10% higher than a lower bound.