Budget-aware scheduling algorithm for scientific workflow applications across multiple clouds. A Mathematical Optimization-Based Approach

Scientific workflows have become a prevailing means of achieving significant scientific advances at an ever-increasing rate. Scheduling mechanisms and approaches are vital to automating these large-scale scientific workflows efficiently. On the other hand, with the advent of cloud computing and its easier availability and lower cost of use, more attention has been paid to the execution and scheduling of scientific workflows in this new paradigm environment. For scheduling large-scale workflows, a multi-cloud environment will typically have a more significant advantage in various computing resources than a single cloud provider. Also, the scheduling makespan and cost can be reduced if the computing resources are used optimally in a multi-cloud environment. Accordingly, this thesis addressed the problem of scientific workflow scheduling in the multi-cloud environment under budget constraints to minimize associated makespan. Furthermore, this study tries to minimize costs, including fees for running VMs and data transfer, minimize the data transfer time, and fulfill budget and resource constraints in the multi-clouds scenario. To this end, we proposed Mixed-Integer Linear Programming (MILP) models that can be solved in a reasonable time by available solvers. We divided the workflow tasks into small segments, distributed them among VMs with multi-vCPU, and formulated them in mathematical programming. In the proposed mathematical model, the objective of a problem and real and physical constraints or restrictions are formulated using exact mathematical functions. We analyzed the treatment of optimal makespan under variations in budget, workflow size, and different segment sizes. The evaluation's results signify that our proposed approach has achieved logical and expected results in meeting the set objectives