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

This article introduces how to deploy GPU to meet the server workload requirements, the content is very detailed, interested friends can refer to, hope to be helpful to you.

With the change of computing requirements in modern data centers, high-performance GPU has become an important part of server virtualization. Server loads, such as big data analysis, require higher visualization and presentation capabilities for data to display more complex data. The server also needs more graphics processing power to meet the requirements of the virtualized terminal tasks.

So far, most server vendors have ignored graphics requirements because traditional workloads such as databases or files and AD servers do not require such deductive and visual capabilities. The system designer chose to abandon GPU to meet the requirements of reducing server costs and minimizing system energy consumption.

However, with the advent of the era of virtualization and the increasing demand for multimedia and visualization tools, business units are prompted to reconsider the graphics display hardware of the server. With the development of server technology, manufacturers begin to consider providing servers with hardware directly integrated with GPU.

Be sure to plan and test before deploying a high-performance GPU, because servers, unlike PC and workstations, do not provide slot space and power cables.

The role of high-performance GPU

The GPU in the server plays the same role as in the client computer: GPU frees the application's graphics processing instructions from the main processor. This process releases the graphics processing instructions of the main processor resources for other tasks and executes applications on hardware, meeting our current needs for more sophisticated, lifelike rendering, video processing, and visualization. Without GPU, graphics instructions require software emulation and then consume the resources of the main processor, resulting in unbearably low performance.

For example, application virtualization can allow servers to support applications shared by multiple users. If the shared application requires a graphics feature, such as a video rendering tool, the server must provide that functionality. In addition, virtual desktops (VDI) can also support terminal centralization on the server. In this case, 3D modeling software and other graphics tools, which usually run in a desktop PC environment, are now on server-based virtual machines, requiring additional graphics processing capabilities as well.

Install GPU on a virtual server

High-performance GPU is usually installed in traditional servers in the form of highly specialized graphics adapter cards, such as NVIDIA's Tesla, in the server's available PCIe slot. This is the easiest and most common way to improve existing servers without onboard GPU, but there are still some challenges to consider.

These GPU cards are usually large and power-demanding devices, while servers usually provide only one or two PCIe slots, one of which may have been occupied by other PCIe extension devices, such as multi-port network cards or IO accelerators. Even if there is a suitable free slot, a GPU card with a complete large heat sink and fan usually exceeds the size of the available space.

You need to keep in mind that GPU cards require hundreds of watts of system power supply. This requirement may cause problems for server platforms with smaller power supply modules to achieve high efficiency and energy consumption. Some systems may need to upgrade the power supply and provide additional power cables to accommodate the GPU card. The PCIe bus cannot provide so much power to the supported devices.

Adding GPU cards is always a project that needs to be verified. IT experts will need to evaluate the deployment of GPU cards very carefully and verify the capabilities of the server to determine whether it can support the GPU load under various operating conditions.

In any case, the new server has directly considered the GPU problem in the server's motherboard design. Ultra Micro's SuperServer 1027GR-TRFT, combined with the G200eW GPU of on-board Matrox, makes your integration easier. The GPU does not require PCIe slots, and the power supply has taken into account the need to run additional GPU chips.

Software requirements of server GPU

For any computer, the graphics display platform is a necessary subsystem, but in addition to physical space and power supply problems, GPU also needs to be compatible with the server's operating system.

NVIDIA currently supports only 32-bit and 64-bit Linux for servers. In order to expand its scope of application, the GPU needs drivers that support Windows Server2012, including support for hypervisor, such as vSphere or Hyper-V. In short, mechanisms are needed to ensure that CPU cores can share GPU. This is especially important for VDI deployments where many desktops require graphics.

For decades, server vendors have avoided using graphics features, preferring to put high-performance graphics rendering and visualization on end devices with separate card systems. As virtualization enables applications and terminals to be integrated into the data center, graphics capabilities also need to be migrated to server hardware. IT experts need careful planning to avoid storage or power bottlenecks and compatibility issues when retrofitting enterprise-class GPU to existing servers.

On how to deploy GPU to meet the server workload needs to share here, I hope the above content can be of some help to you, can learn more knowledge. If you think the article is good, you can share it for more people to see.

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.