Algebraic Multigrid and Cloud Computing: Enhancing Scalability and Performance
Autour(s)
- Kubura Motalo, Lolade Nojeem, Joe Ewani, Atora Opuiyo, Ibrina Browndi
Abstract
Algebraic Multigrid (AMG) is a powerful computational technique used in scientific computing to solve linear systems of equations quickly and efficiently. With the rise of cloud computing, researchers and practitioners are exploring ways to leverage the power of cloud platforms to improve the scalability and performance of AMG. This article provides an overview of AMG, its benefits, and its limitations in cloud computing environments. Additionally, the article explores the recent developments in cloud-based AMG algorithms and parallel computing techniques to enhance scalability and performance. Algebraic Multigrid (AMG) is a powerful computational technique used in computer science to solve linear systems of equations quickly and efficiently. This article provides an in-depth review of AMG, including its principles, and current state-of-the-art techniques. Additionally, the article explores the benefits of combining AMG with cloud computing, particularly with respect to improving performance and scalability. The literature review reveals that the use of cloud computing with AMG has shown promising results, particularly in scientific simulations and other computationally intensive applications. Algebraic multigrid (AMG) is a powerful preconditioner for solving large-scale linear and nonlinear problems in computational science and engineering. However, the scalability and performance of AMG can be limited by the hardware and software environments, especially in cloud computing. In this paper, we investigate the enhancement of AMG scalability and performance in cloud computing environments by analyzing the impact of various factors, such as communication overhead, load balancing, and data locality. We propose a novel parallel algorithm for AMG that takes advantage of the cloud computing resources and optimizes the communication and computation balance. We demonstrate the effectiveness and efficiency of our approach by conducting a series of experiments on different cloud platforms and problem sizes. The results show that our approach can significantly improve the scalability and performance of AMG in cloud computing environments.