Integrating Generative AI in Cloud Computing Architectures: Transformative Impacts on Efficiency and Innovation

Abstract

To cope with the ever-increasing information processing and enable the developing machinery of human civilization to grow smarter, it becomes necessary to build a cloud computing environment that can process large amounts of data at high speeds and low costs. For neural network-based inference applications, GPUs are commonly used. It is for this reason that we believe large-scale GPU clusters, which are currently indispensable in cloud computing processing, can be expected to solve the problem that they are highly efficient for neural network processing, yet less efficient for other processing methods, including basic users. In this study, we propose deep neural network-based processing technology that can utilize the technology for various processing by converting new processing applications, which are not only neural network-based inference, into multiple tasks that can be executed by GPUs. By making them executable, even conventional data distribution, effective job allocation, etc. can become much easier to achieve, resulting in synergistic improvements in GPU cluster utilization.In addition, users using a cloud that integrates the proposed method can minimize processing costs even when neural network processing and other processing methods are mixed and used, and can spend the budget obtained from that cost savings on realizing new business innovations. It is expected that this approach will profoundly influence the direction of cloud computing business deployments.