Scheduling Workflows in Multi-Tenant Cloud Computing Environments

In the last few years, software deployment extensively moved in a cloud environment. Following this popularity, different solutions based on Infrastructure-as-a-Service (IaaS) appeared, and one of them is Workflow-as-a-Service (WaaS) paradigm. Such platforms allow users to upload their workflows described as directed acyclic graphs and delegate responsibility for managing and executing tasks on virtual machines (VMs). This type of environment is popular among scientists, as their workflows are usually too large in structure for manual distribution of tasks and control of data flow among VMs. Nowadays, there are many different algorithms for scheduling workflows, and all of them were evaluated in various setups and environments. Therefore, this work aims to compare some of the scheduling algorithms comprehensively. All studied algorithms optimize the execution process of a workflow within a user-specified constraint —— minimize either the execution time within a budget or the execution cost within a deadline. The comparison should be conducted in a unified environment to get accurate results. The best solution is to use a cloud simulator. However, it was decided not to use the existing ones, as most of them are either outdated or too complicated for the aforementioned purposes. Thus, a simulator should be implemented. A simple cloud simulator would be sufficient for this work to implement models considered by the authors of the studied algorithms. The main purpose of this work is to select relevant and the most promising algorithms and implement and evaluate them in identical conditions with providing a comprehensive overview of their performance.