A method to solve the network performance bottleneck in large-scale video cloud applications

Abstract

In order to overcome the shortage of virtual technologies in high I / O load application scenarios, the invention provides a method for solving network performance bottlenecks in large-scale video cloud applications. The method includes hardware environment implementation steps and software environment implementation steps, wherein the hardware environment implementation steps are as follows: a PC server that supports an SR-IOV function is configured, and the PC server is provided with a network card that supports the SR-IOV function; and the software environment implementation steps are as follows: a virtualized underlying layer of the PC server uses an XenServer platform; the SR-IOV function is started in BIOS settings of the PC server and is adjusted to an SR-IOV mode in the BIOS settings of the network card; and the startup parameters of the XenServer are modified, and the network card directly used by a video application virtual machine needs to be configured. According to the method provided by the invention, a physical network card is directly connected to the video application virtual machine for use based on an SR-IOV technology, a video flow processing program can be directly interacted with the physical network card, a virtualization management layer and a virtual switch layer are skipped, and thus the problem of network packet loss and delay jitter in virtualized environments can be solved.

Description

technical field [0001] The present invention relates to high-performance I / O demand scenarios in the field of computer virtualization, and in particular to a method for solving network performance bottlenecks in large-scale video cloud applications by using SR-IOV technology. Background technique [0002] Since the development of virtualization technology, the main applicable scenarios are concentrated in CPU-intensive and memory-intensive applications, such as WEB applications. And memory virtualization efficiency is high (up to 90%). However, in application scenarios involving high I / O loads, such as large-scale video applications, high-load database applications, 3D games, etc., virtualization technology is rarely used because of the poor performance of I / O virtualization ( 30%-60%), which also means that these applications cannot be virtualized and cannot enjoy the various benefits brought by the virtualization architecture, such as rapid deployment, fast migration, har...

AI Technical Summary

Problems solved by technology

However, in application scenarios involving high I / O loads, such as large-scale video applications, high-load database applications, 3D games, etc., virtualization technology is rarely used because of the poor performance of I / O virtualization ( 30%-60%), which also means that these applications cannot be virtualized and cannot enjoy the various benefits brought by the virtualization architecture, such as rapid deployment, fast migration, hardware decoupling, cluster HA, etc.

[0003] In the traditional virtualization technology architecture, since all network traffic in the virtual machine must be transferred through the underlying virtualization management program and virtual switch, when the network traffic is large (greater than 400Mbps), it will cause a huge amount of CPU interruption and CPU interruption. Context switching, in severe cases, will occupy all CPU resources, resulting in packet loss, large delay jitter and other faults in the virtual machine network

Images