Shulou(Shulou.com)05/31 Report--

ZStack GPU solution is what kind of, many novices are not very clear about this, in order to help you solve this problem, the following editor will explain for you in detail, people with this need can come to learn, I hope you can gain something.

In the previously released version 3.5.0 of ZStack, ZStack added support for NVIDIA and AMD vGPU solutions to implement a complete GPU cloud platform solution that supports GPU transparent transmission and vGPU GPU solutions in the cloud, which can help users to achieve richer GPU application scenarios such as industrial design, graphics rendering, teaching and research, and deep learning in the cloud.

Preface

With the rapid development of virtualization and cloud computing technology, more and more users migrate applications to the cloud. In this process, users hope that some specific business scenarios can not only obtain flexible and efficient resource delivery capabilities in the cloud environment, but also obtain the same resource computing performance as physical machines. GPU as a resource equipment, in recent years, with the deepening of deep learning related applications such as big data and artificial intelligence, the application demand of this technology in the cloud is deepening day by day.

The purpose of this article is to give readers an overview of GPU solutions in the cloud in an easy-to-understand way.

I. get to know GPU

Before we can understand GPU, we need to understand the difference between GPU and CPU, which we are most familiar with and commonly used: although both are designed to accomplish computing tasks, they each have their own focus.

CPU consists of several cores designed specifically for serial sequential execution of instructions, mainly for general computing scenarios

8 computing cores, each of which has relatively strong computing power

GPU, on the other hand, has a massive parallel computing architecture consisting of thousands of smaller cores. More emphasis on parallel computing scenarios

There are 1000 computing cores, the computing power of a single core is relatively general, and the comprehensive computing power is strong.

Give an example that is easier to understand:

CPU can be compared to a math professor, who is more efficient when solving one or more difficult math problems.

GPU can be compared to 100 middle school students who are more efficient when solving hundreds of low-difficulty math problems.

Therefore, in practical application scenarios, GPU is very suitable for scenarios that require a lot of computing power such as simplicity, repetition, parallelism, matrix and so on. Here are two typical GPU application scenarios to illustrate the requirements.

2. GPU application scenario 1. Heavy GPU computing power requirement scenario

The traffic management department of a certain city built the big data intelligent traffic dispatching system to alleviate the high-speed traffic congestion around the city, and the construction plan built the system in the cloud environment to solve the problems such as long installation cycle and low efficiency of operation and maintenance when deployed in the physical machine environment. Therefore, how to deliver the physical GPU equipment to the artificial intelligence application deployed in the cloud environment for use is the key technical requirement of the project.

two。 Scenario of Light GPU Computing Power requirement

A manufacturing enterprise expects to deploy cloud desktops for industrial designers to protect enterprise data security and simplify desktop operation and maintenance management. But this needs to rely on the graphical computing power of GPU to meet the need to run industrial design software in a cloud environment.

In the above scenarios, the delivery of GPU resources in the cloud through GPU transmission and vGPU technology is the main technical means to meet the needs of users.

Let's further introduce the above two cloud GPU implementations.

3. GPU cloud solution 1. GPU transparent transmission

The main technical principle of GPU transparent transmission is that by mapping the PCI memory address of the GPU device on the host (physical machine) to the CVM, it can bypass the virtualization manager in the cloud environment and load the GPU device directly to the CVM for use.

After transparent transmission, the advantages and disadvantages of using GPU devices on physical machines are as follows:

Advantages: performance loss

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.