Scheduling algorithms for heterogeneous fog networks

Abstract

Due to the significant communication delay to user tasks, the cloud is not ideal for executing real-time tasks with stringent deadlines. Edge computing has been successful in attracting significant attention from both industry and academia. Fog/Edge computing extends the functionality of the traditional cloud data center (cdc) using micro data centers (mdcs) located at the edge of the network. These mdcs provide both computation and storage to applications. Their proximity to users makes them a viable option for executing jobs with tight deadlines and latency constraints. Moreover, it may be the case that these mdcs have diverse execution capacities, i.e. they have heterogeneous architectures. The implication for this is that tasks may have variable execution time on different mdcs. In one of our works, we propose PASHE (Privacy Aware Scheduling in a Heterogeneous Fog Environment), an algorithm that schedules privacy constrained real-time jobs on heterogeneous mdcs and the cdc. In order to model the security/privacy constraints of applications, three categories of tasks have been considered: private, semi-private and public. Private tasks with tight deadlines are executed on the local mdc of the users. Semi-private tasks with tight deadlines are executed on “preferred” remote mdcs. Public tasks with loose deadlines are sent to the cdc for execution. We also take account of user mobility across different mdcs. If the mobility pattern of users is predictable, PASHE reserves computation resources on remote mdcs for job execution. In another work, we propose algorithms that schedule a set of real-time tasks on a fog-cloud architecture. We consider three types of tasks - hard, firm and soft. The execution framework consists of three kinds of processors - embedded processors, fog processors and cloud processors. Tasks are scheduled on appropriate processors based on their deadline requirements. In general, hard real-time tasks are executed on the embedded processors, firm real-time tasks on the edge processors, and soft real-time tasks on the cloud processors. We also propose a sufficient schedulability condition for these tasks on this three-tier architecture. Simulation results show that our proposed approaches offer superior performance versus other scheduling algorithms in a fog computing environment, taking account of application real-time behaviours, mdc heterogeneity, user mobility and application security/privacy.