Future Generation Computer Systems ( IF 7.5 ) Pub Date : 2021-01-08 , DOI: 10.1016/j.future.2021.01.003 A. Asensio , X. Masip-Bruin , J. Garcia , S. Sánchez
Smart environments rely on the cloud for most computation activities; however, leveraging the availability of resources at the edge could complement the cloud capabilities to provide a resources continuum spectrum that takes advantage of the benefits of both technologies, cloud and edge computing. In this scenario, applications are usually decomposed for execution in sets of tasks, which are in turn encapsulated into virtual components such as containers. Containers are lightweight implementations of virtual machines, improving efficiency and portability in distributed applications. Traditional containers scheduling in cloud is a well-known problem, but when the environment is heterogeneous, as it is in the scope of edge computing and edge-cloud systems, the problem becomes more challenging. In this paper we present the Concurrent Container Clusters Scheduling problem (C3S) aimed at optimizing the problem of placing containers in clusters of heterogeneous nodes satisfying a set of resource requirements, quality of service limitations, and considering additional stringent constraints in terms of applications execution in isolation for security guaranteeing. The C3S problem has been formulated using Integer Linear Programming with the dual objective of minimizing the number of applications rejected while minimizing the number of nodes used for computation. We have evaluated the optimality of this approach, analyzed the performance in terms of solving time and, finally, created a heuristic approach to solve the problem in realistic high demanding scenarios.
中文翻译:
异构智能环境下并发容器集群调度的最优性
智能环境大多数情况下都依靠云计算。但是,利用边缘资源的可用性可以补充云功能,以提供资源连续谱,从而充分利用云和边缘计算技术的优势。在这种情况下,通常将应用程序分解为在一组任务中执行,这些任务又被封装到诸如容器之类的虚拟组件中。容器是虚拟机的轻量级实现,可提高分布式应用程序的效率和可移植性。传统的容器在云中进行调度是一个众所周知的问题,但是当环境是异构的时,因为它在边缘计算和边缘云系统的范围内,这个问题就变得更具挑战性。在本文中,我们提出了并发容器集群调度问题(C3S),旨在优化将容器放置在满足一组资源要求,服务质量限制并在应用程序执行方面考虑其他严格约束的异构节点集群中的问题。隔离以确保安全。C3S问题是使用整数线性规划公式化的,其双重目标是最大程度地减少拒绝的应用程序数量,同时最小化用于计算的节点数量。我们评估了这种方法的最优性,分析了解决时间方面的性能,最后创建了一种启发式方法来解决现实的高要求场景中的问题。