Method for automatically managing and controlling virtual GPU (Graphics Processing Unit) resource in cloud gaming

Abstract

The invention provides a method for automatically managing and controlling a virtual (Graphics Processing Unit) resource in cloud gaming. The method comprises the steps of establishing an Auto-vGPU framework; selecting key index data as system input by using an LASSO (Least Absolute Shrinkage and Selection Operator)-based dimensionality reduction algorithm for supporting an input and output model, which fits the low dimension, of the Auto-vGPU framework; automatically controlling the parameter configuration in a vGASA module in the Auto-vGPU framework by utilizing a parameter automatic configuration algorithm of a PI (proportional-Integral) controller, reducing the manual operation, and supporting the Auto-vGPU framework to keep good performance in a dynamic and complicated cloud environment. According to the method provided by the invention, the automatic management on the virtual GPU resource is realized, an LASSO / LARS (Least Angle Regression) dimensionality reduction technology is used, the input capacity of each game in a virtual machine is automatically reduced, and an online controller is also designed through adopting a method for automatically configuring PI parameters according to expected performance.

Description

technical field [0001] The present invention relates to a virtual GPU management and control method, in particular to a virtual GPU resource autonomous management and control method in a cloud game. Background technique [0002] Cloud gaming is a gaming method based on cloud computing. In the operating mode of cloud gaming, all games run on the server side, and the rendered game screen is compressed and sent to users through the network. In the concept of cloud games, the player's game client is regarded as a thin client. Players can still play new games conveniently without high-end processors and graphics cards. Cloud games can rely on reasonable resource allocation to maximize economic benefits, but this requires solving resource scheduling and management problems. [0003] Currently, there is an adaptive virtual GPU resource scheduling scheme vGASA. The solution runs in a virtualized environment and uses GPU virtualization technology to share resources. One of the cha...

AI Technical Summary

Problems solved by technology

[0006] (2) Unknown changes may lead to system performance degradation, which will require the entire system (including models and controllers) to be redesigned;

[0007] (3) In a comprehensive virtualization environment, manual offline design will spend a lot of time on the daily operation of a large number of hosts and virtual machines;

[0008] (4) A given controller can only play a role when the configurations of the host and the virtual machine are specific and unchanged, and in the online cloud game platform, the dynamic cloud environment phenomenon of virtual machine migration often occurs

Images