VMware HorizonView 5.2 3D图形性能深度剖析

需积分: 7 83 浏览量更新于2024-07-25 收藏 687KB PDF 举报

本篇技术白皮书主要探讨了VMware HorizonView 5.2与硬件加速3D图形性能的研究报告。随着虚拟桌面基础设施（VDI）的广泛应用和带宽成本的降低，企业不再受限于低分辨率桌面，而是寻求提供更高质量的图形体验，特别是在支持3D密集型工作负载时。本报告详细介绍了以下内容： 1. **硬件加速3D图形**：重点在于如何利用硬件加速技术来提升VMware HorizonView 5.2中3D桌面的性能，包括支持的配置和最佳实践。通过硬件加速，系统能够更有效地处理复杂的3D图形任务，如CAD应用和游戏。 2. **3D桌面在View中的应用**：讨论了如何在视图中实现流畅的3D桌面环境，以及针对不同类型工作负载（如轻量级3D任务、交互式用户体验和CAD应用如SolidWorks和SolidEdge）的优化策略。 3. **支持的配置**：列出了满足硬件加速3D图形需求的理想硬件配置，包括处理器、显卡和内存规格等，确保用户获得最佳性能。 4. **最佳实践**：提供了优化资源共享、最大化VM consolidation（虚拟机整合）以及跟踪计算资源的实用建议，帮助管理员充分利用硬件资源。 5. **客户端大小和vSGA性能测试**：探讨了不同配置下客户端的内存需求和vSphere Graphics Accelerator (vSGA) 的性能表现，这对于确保客户端性能至关重要。 6. **测试床架构**：概述了用于性能测试的实验环境，包括服务器、网络和客户端组件的配置，以保证结果的可靠性。 7. **性能指标**：报告中包含了衡量硬件加速3D图形性能的关键指标，如帧率、延迟和吞吐量，以便量化和比较不同场景下的性能差异。 8. **可扩展性结果**：研究了随着更多用户和更复杂的工作负载，系统在性能上的表现，以及如何应对规模增长的需求。 9. **特定工作负载性能**：分别分析了针对不同3D工作负载（如轻量级图形和复杂的CAD应用）的实际性能数据，以帮助企业评估是否能满足其业务需求。 10. **结论**：总结了整个研究的主要发现，以及对VMware HorizonView 5.2硬件加速3D图形技术在实际部署中的价值和推荐。 11. **参考文献**：提供了进一步阅读和研究的相关资料来源，供读者深入探究。通过阅读这篇报告，企业可以更好地了解如何在VMware HorizonView环境中利用硬件加速3D图形技术，以提高工作效率并支持更复杂的图形密集型应用。

VMware Horizon View 5.2 and

Hardware Accelerated 3D Graphics

TECHNICAL WHITE PAPER /4

Best Practices

View 5.2 and PCoIP dynamically adapt to the available CPU and bandwidth resources to present the optimal user

experience. Even when tens of VMs are sharing a single physical GPU, vSphere ensures that the resource is fairly

shared between the different VMs. As a result, very little out-of-the-box configuration is required deliver peak

performance:

•

Configure VMs to use VMXNET3 NICs.

•

Consider disabling PCoIP’s build-to-lossless mode.

•

Set the MaxAppFrameRate (see Optimizing Resource Sharing).

The next sections describe additional considerations for maximizing performance and efficiency of a 3D desktop

workload.

Maximizing VM Consolidation

The administrator console provides the ability to configure the amount of graphics or video RAM (termed VRAM)

allocated to each VM. The default per-VM VRAM allocation is 128MB. While increasing the per-VM VRAM

allocation might deliver higher performance, it will limit the number of VMs that can simultaneously share a GPU;

the GPU’s memory is sub-divided between the VMs. Accordingly, if the GPU has 4GB of memory, and the VMs are

configured to use the default 128MB of VRAM, it is possible for up to 32 VMs to simultaneously share the GPU.

Similarly, if the VMs are configured to use 512MB each, the per-GPU consolidation ratio may be reduced to 8. To

further increase the number of VMs supported per server, vSphere 5.1 can support multiple GPUs per system.

For Aero and basic 3D applications, use of the default VRAM allocation is sufficient and is recommended to

ensure maximum consolidation ratios. Higher allocations can be reserved for VMs that will run more intensive 3D

operations, where additional GPU resources will deliver improved performance.

For pools containing more VMs than can be simultaneously supported by the GPU, additional VMs will not boot

once the GPU resources have been exhausted. In this situation, rather than explicitly setting the pool to use

hardware 3D in the administrator console, it is advantageous to use the "automatic" option; this enables any

additional VMs that might be needed, over and above the GPU’s capabilities, to be supported using vSphere's

software renderer solution.

In contrast, for situations where two distinct groups of users share a server—one group requiring hardware 3D

acceleration and one not—it is best to configure these two groups of VMs as separate pools. Use the

administrator console to explicitly configure the hardware group to use hardware accelerated 3D and the other

group to use either software 3D or even no 3D, as appropriate. Changes to a pool’s 3D strategy are handling

automatically by View and vSphere and do not require manual configuration of the desktop VMs.

Optimizing Resource Sharing

In contrast to a physical workstation that has sole use of its GPU, in the virtualized environment GPUs become a

shared resource. As a result, it is important to ensure that each VM does not use the GPU resource in a wasteful

manner. For instance, in many situations it often does not make sense for a 3D application to render hundreds of

frames per second if View is configured to remote at a lower frame-rate (30fps is the default setting). For these

situations, View provides a registry setting to limit the maximum application frame rate. This can either be

configured in the template VM or on a per-VM basis, and the value should typically be set to the maximum frame

rate that is being used by PCoIP. This configuration is achieved by using the following registry setting

(REG_DWORD):

HKLM\SOFTWARE\VMware, Inc.\VMware SVGA DevTap\MaxAppFrameRate

Setting this registry entry for a 3D workload has been found to significantly improve the performance and

consolidation ratios achievable.

剩余16页未读，继续阅读

popjoker

粉丝: 0
资源: 1

VMware HorizonView 5.2 3D图形性能深度剖析

移动设备3D图形：OpenGL ES与M3G应用

游戏编程中的3D图形学：纹理，模型与采样深度解析

C# ArcGIS实现3D Graphics与曲线距离测量方法

程序员考试刷题-udacity_interactive3dGraphics:udacity_interactive3dGraphics

3DGraphics302项目

3Dgraphics.exe

MagneticField_3DGraphics(Czechic Comments)

IntroTo3DGraphics:快速入门 Blender 和 Blender API

graphics-algorithm：3D图形学算法代码。包括软渲染，光线追踪，PBR等〜

JavaPrintService-Java打印API-用户手册-中文版

最新资源