Author(s)

Zhen Li

Corresponding Author

Zhen Li

ABSTRACT

GPU is a new commercial computing model and an extension and expansion of technologies such as distributed parallel processing and grid computing. It represents a new stage in the development of parallel computing technology and belongs to a brand new product. With the maturity of this field and the continuous growth of the internet and global enterprises, it is expected that more cloud service providers will emerge. This article focuses on the research and development scenarios of AI(Artificial Intelligence) algorithms and conducts a series of scheduling strategy studies based on GPU to improve the task execution efficiency and cluster resource utilization of the AI algorithm development platform. Finally, the research results indicate that when the historical time slot changes from 50 to 590, the accuracy of prediction becomes more and more accurate as the size of the time slot increases, and the prediction curve gradually becomes flat at 410 to 460. It can be inferred that there may be an extreme value around the time slot size of 460, and exceeding this extreme value has little impact on predicting low utilization time results. During the process of increasing average execution power, resource correlation algorithms are used to schedule tasks, reducing the overhead of the cloud system data center network and improving the system's resource utilization.

KEYWORDS

Artificial intelligence algorithms; GPU; Resource optimization and scheduling; Control; Application