3D FlexPod: 3D Virtual Desktop Infrastructure Visualization with TGX on FlexPod Express or FlexPod Datacenter

Transcription

1 Technical Report 3D FlexPod: 3D Virtual Desktop Infrastructure Visualization with TGX on FlexPod Express or FlexPod Datacenter Chris Rodriguez, NetApp Mark Hirst, Mechdyne March 2017 TR-4576 In partnership with

2 TABLE OF CONTENTS 1 Executive Summary Reference Architecture Objectives Introduction Outstanding Performance Solution Summary Solution Configuration NVIDIA GPU Cards and NetApp FlexPod Installation and Configuration TGX TGX Windows Installation TGX Windows Installation TGX Linux Installation TGX Mac Installation (Receiver Only) Documentation Links Performance Optimization Latency Technical Features Encoding and Decoding IT Management and Security Conclusion Use Cases Sizing Summary References Recognition Authors Acknowledgements LIST OF TABLES Table 1) Cisco Nexus 9372-A cabling information Table 2) Cisco Nexus 9372-B cabling information Table 3) NetApp Controller-1 cabling information Table 4) NetApp Controller 2 cabling information Table 5) Cisco UCS fabric interconnect A cabling information Table 6) Cisco UCS fabric interconnect B cabling information D FlexPod: 3D VDI Visualization with TGX on

3 Table 7) Hardware and drivers for TGX Windows installation Table 8) Hardware and drivers for TGX Linux installation Table 9) Bandwidth consumption Table 10) Sizing information LIST OF FIGURES Figure 1) Ramp phase response time for a NetApp A700 AFF storage controller....6 Figure 2) Component layers for the FlexPod Datacenter with Cisco UCS....9 Figure 3) Dual 3840 x 1600 monitors displaying Schlumberger s Petrel software at 60FPS using TGX with FlexPod. For more information, see the video on 3D FlexPod Figure 4) Communication between the sender and receiver D FlexPod: 3D VDI Visualization with TGX on

4 1 Executive Summary Server virtualization is a mainstream practice in current IT infrastructure deployments. The infrastructure of a virtualized environment typically includes compute nodes (physical servers), switches, and a storage array system. A majority of virtualized environments use shared storage and switches. In addition to the virtualization of standard business applications, video clusters are now making the jump to the virtualized world. This jump has been facilitated by the advent of powerful graphics processing cards (GPUs) and drivers provided by hypervisor companies. There are many reasons why companies are considering moving these video application-related datasets. The following challenges are experienced by companies in a decentralized 3D visualization environment. Dataset Size and Complexity Workflows in the 3D visualization market are continually evolving. Between the rapid growth in dataset size and the need to integrate diverse, multidisciplinary data types, companies are finding more ways to accurately understand their data in 3D visual representation. Companies rely on accurate data representation to interpret their research, and visualizing large, high-resolution datasets is crucial in making key decisions. Requirements for High-Resolution Interpretation There is a growing requirement for large desktop workstations with 4K resolution or higher. The consumer market is driving down the cost of 4K displays. Therefore, demand is increasing for this new technology as users become aware of the additional insights that can be obtained from visualizing their data at higher resolutions. As a result, IT departments are struggling with existing technologies that cannot keep pace or support this change in requirements. Dispersed Workforce Today s workforce is becoming increasingly dispersed. Remote access to secure data, outsourcing, and close collaboration between teams is becoming an important component of 3D visualization requirements. Allowing a mobile workforce wherever it happens to be to visualize graphics pictures is becoming an essential part of many companies requirements. This requires either a local copy of the dataset, which can be difficult and potentially risky to provide, or remote visualization. Up to now, remote visualization solutions have been complicated to deploy and manage and haven t provided the quality or responsiveness that users expect. Distributed Datasets Due to the size and complexity of datasets in 3D visualization, transferring these large files over a network can cause performance issues. This difficulty is caused by the lack of 10Gb connectivity between the desktop and the data center where the data resides. As a result, application performance and the user s productivity can decrease. Moving these large datasets across the network is costly and inefficient, can negatively affect application and network performance, and poses a significant security risk. Locating these datasets on shared storage in a siloed converged infrastructure prevents network saturation and eliminates the issues with distributed datasets. Data Security Many IT organizations have provided either high-powered mobile laptops or workstations to their users. To counter the performance issue of copying large datasets over the network, some users store data locally to save processing time, creating a significant security risk. Sensitive company data under these circumstances can be exposed to theft, data loss, viruses, malware, and ransomware. 4 3D FlexPod: 3D VDI Visualization with TGX on

5 User Experience A key frustration for any type of remote compute environment is the experience of users at their desk. Poor image quality, unresponsive applications, and a lack of support for high-resolution monitors are common complaints that can affect productivity and limit the success of any virtual desktop infrastructure (VDI) deployment. The minimum benchmark for the user experience is that the remote desktop should feel like a local machine at the user s desk. Unfortunately, VDI solutions have historically failed to achieve this target. While many companies have put their workstations on flash, that is only one part of the equation for the user experience. Placing both the application and the data together on flash provides the optimal end-user experience. The Solution Cisco, NetApp, NVIDIA, and Mechdyne in a partnership have created a solution for the 3D visualization market. Cisco and NetApp have created a converged infrastructure that provides all-flash, highperformance storage. NVIDIA provides the high-performance graphics processing cards (GPUs), and Mechdyne provides the TGX protocol for graphics visualization and end-user collaboration. Cisco has developed firmware to support NVIDIA GPU cards in B-series M4 servers and C-Series servers. NVIDIA has developed drivers for the VMware hypervisor that allows the use of NVIDIA GPU cards in a shared VDI environment. The files processed by these VDI video clients are typically very large. NetApp provides superior storage for CIFS access of these large input files. NetApp systems also allow you to use CIFS storage for the TGX VDI software and storage requirements of this environment. Graphics display users tend to be small groups that use a portion of the overall virtualized server environment. The FlexPod Datacenter converged platform is positioned for large enterprise deployments consisting of a mixed virtualized environment that includes NVIDIA GPU users. The capability to use multiple workloads and multiple protocols (including those used by NVIDIA GPU users) differentiates NetApp All Flash FAS (AFF) storage from its competition. Since the testing described in this reference architecture was performed, NetApp refreshed the entire AFF product line so that all new platforms are acceptable in this solution. The NetApp A700 model replaces the NetApp AFF8080 model. The graph in Figure 1 is from the SPC-1 benchmark testing for the NetApp A D FlexPod: 3D VDI Visualization with TGX on

6 Figure 1) Ramp phase response time for a NetApp A700 AFF storage controller. The storage latency (response time) is less than 0.8ms for over 200,000 IOPS. For more information about the SPC-1 performance benchmark on NetApp A700 storage, see the SPC-1 benchmark report. The FlexPod Datacenter configuration in this reference architecture is composed of the Cisco Unified Computing System (Cisco UCS) mini B-series servers, Cisco Nexus switches, and NetApp AFF8080 series storage with two shelves of solid-state drives (SSDs) (48 SSDs total). FlexPod Datacenter supports the following: A larger user deployment for TGX users The use of NAS volumes (CIFS and NFS) The processing of large files Server performance provided by drivers that support the NVIDIA GPU M6 offload graphics card contained in this reference architecture The TGX remote desktop by Mechdyne optimizes centralized compute, cloud environments, and VDI. TGX supports 4K resolutions and higher, making it the most powerful remote desktop available on the market today. It delivers graphics-intensive data from a remote location in high resolution without sacrificing image quality or the user experience. In addition, TGX maintains storage, compute, network, and remote visualization standards for graphics-intensive applications on both Microsoft and Linux platforms. This improved visual output and control provide the most demanding users with the confidence to make critical decisions based on what they see on the desktop, while simultaneously satisfying IT requirements. TGX provides the following features: Exceptional color accuracy Extreme resolutions, exceeding two times 4K 6 3D FlexPod: 3D VDI Visualization with TGX on

7 High-quality imagery at 60 FPS No image artifacts, including tearing or loss of pixels Display resolution autodetection Remote collaboration Client support for Mac, Windows, and Linux As a result, users experience improved decision making, optimized productivity, and increased collaboration. Benefits to the IT team include increased data security and reduced network effects. NetApp, Cisco, NVIDIA, and TGX have together created the 3D FlexPod integrated solution, which meets the demands of 3D visualization and graphics. 1.1 Reference Architecture Objectives This reference architecture describes how to configure the Mechdyne TGX protocol with NetApp storage and NVIDIA GPU cards in a FlexPod Datacenter configuration with Cisco UCS 5108 B-Series servers and the NetApp AFF8080. This reference architecture uses VMware vsphere 6.0u1 on a Cisco UCS Classic with B200 M4 blade servers, two Cisco Nexus 9000 series switches, and a NetApp AFF8080. The system operates under a FlexPod Datacenter converged infrastructure umbrella. This technical report is a guide to installing and configuring a 3D FlexPod system. Therefore, we assume that the reader is familiar with the architecture, design, infrastructure, and configuration of a FlexPod Datacenter system with VMware ESXi or XenServer. To learn more about the base infrastructure installation, see the relevant Cisco Validated Design (CVD) document. This CVD covers Cisco UCS servers, Cisco Nexus 6300 fabric interconnects, Cisco Nexus 9000 switches, and the NetApp AFF8000 series FlexPod Datacenter product line. In addition, this technical report covers the TGX software installation and configuration on a FlexPod Datacenter configuration. The hardware and software infrastructure described in this technical report was used to conduct the validation and to create this technical report. This proven infrastructure provides step-by-step instructions on configuring TGX software with NVIDIA GPU cards in a FlexPod Datacenter configuration. 2 Introduction Built on more than six years of innovation, FlexPod has evolved to meet the changing needs of our customers and has helped drive their success. Specifically, the FlexPod Datacenter solution provides a rich set of data management features and clustering for scale-out, operational efficiency, and nondisruptive operations. This combination offers you one of the most compelling value propositions in the industry. The IT landscape is undergoing a fundamental shift to IT as a service, a model that requires a pool of compute, network, and storage to serve a wide range of applications and deliver a wide range of services. Innovations such as the FlexPod Datacenter configuration are fueling this transformation. 2.1 Outstanding Performance For the FlexPod Datacenter configuration, we used an AFF8080 with 48 SSDs in two DS2246 shelves (24 drives per shelf). This configuration provides outstanding performance for CIFS and for 3D visualization applications that requires large data file access. The AFF8080 is a superior storage array for a mixed workload enterprise, and it also provides ample performance for a VDI environment with less than 1ms latency. The NetApp AFF8080, when matched with a Cisco UCS converged infrastructure with Cisco Nexus data switches, provides the performance required for large-scale NVIDIA GPU environments. 7 3D FlexPod: 3D VDI Visualization with TGX on

8 2.2 Solution Summary This solution is a converged infrastructure with full redundancy that is based on the following hardware and software: FlexPod Datacenter: One NetApp AFF8080HA two-node storage cluster with 48 SSDs Two Cisco Nexus 9372PX switches for data One Cisco UCS Classic 5108 chassis Two 6248UP fabric interconnects One NVIDIA GPU M6 card VMware vsphere ESXi 6.0 vcenter Server Appliance 6.0 Windows 2012 Server for the virtual machines (VMs) NetApp clustered Data ONTAP NetApp Virtual Storage Console (VSC) 6.2 plug-in for vcenter 6.0 Windows 7 for VDI desktops Windows 10 for VDI desktops TGX Solution Configuration This FlexPod Datacenter configuration is based on a best practices design for TGX on a vsphere 6.0 hypervisor running on a FlexPod Datacenter architecture with NVIDIA GPU cards. The base infrastructure configuration was tested successfully with NVIDIA M6 cards running in a rock-solid FlexPod Datacenter infrastructure architecture. This section provides the configuration details for the FlexPod Datacenter infrastructure with the NVIDIA M6 cards. Figure 2 displays the hardware components and architectural design used in this reference architecture. 8 3D FlexPod: 3D VDI Visualization with TGX on

9 Figure 2) Component layers for FlexPod Datacenter with Cisco UCS. This architecture is divided into three distinct layers: The Cisco UCS compute platform The network access layer and the LAN Storage access to the NetApp AFF8080 EX The solution contains the following hardware: Two Cisco Nexus 9372PX layer 2 access switches Four Cisco UCS 5108 blade server chassis with two built-in UCS-IOM-2208XP I/O modules Two Cisco UCS B200 M4 blade servers with Intel Xeon E5-2660v3 2.6GHz 10-core processors, 128GB of RAM at 2133MHz, and VIC1340 mezzanine cards for the hosted infrastructure with N+1 server fault tolerance 9 3D FlexPod: 3D VDI Visualization with TGX on

10 Twenty-six Cisco UCS B200 M4 blade servers with Intel Xeon E5-2680v3 2.5-GHz 12-core processors, 384GB of RAM at 1866MHz, and VIC1340 mezzanine cards for the virtual desktop workloads with N+1 server fault tolerance NetApp AFF8080EX-A dual controller storage system and two DS2246 disk shelves with 48x 800GB SSDs and 10GbE ports for iscsi and NFS/CIFS connectivity Sixteen Cisco UCS B200M3 blade servers with Intel E v2 processors, 256GB of RAM, and VIC1240 mezzanine cards plus a NetApp FAS2240 for the Login VSI launcher and logging infrastructure (the test rig is not part of the solution) Table 1 through Table 6 provide the details of all the connections in use. Table 1) Cisco Nexus 9372-A cabling information. Local Device Local Port Connection Remote Device Remote Port Cisco Nexus 9372 A Eth1/1 10GbE NetApp controller 2 e0e Eth1/2 10GbE NetApp controller 2 e1a Eth1/3 10GbE NetApp controller 1 e0e Eth1/4 10GbE NetApp controller 1 e4a Eth1/5 10GbE NetApp controller 2 e0f Eth1/6 10GbE NetApp controller 2 e4a Eth1/7 10GbE NetApp controller 1 e0f Eth1/8 10GbE NetApp controller 1 e1a Eth1/17 10GbE Cisco UCS fabric interconnect A Eth2/1 Eth1/18 10GbE Cisco UCS fabric interconnect A Eth2/2 Eth1/19 10GbE Cisco UCS fabric interconnect B Eth2/3 Eth1/20 10GbE Cisco UCS fabric interconnect B Eth2/4 Eth1/49 40GbE Cisco Nexus 9372 B Eth1/49 Eth1/50 40GbE Cisco Nexus 9372 B Eth1/50 MGMT0 GbE GbE management switch Any Note: For devices requiring GbE connectivity, use the GbE copper SFP+s (GLC-T=). Table 2) Cisco Nexus 9372-B cabling information. Local Device Local Port Connection Remote Device Remote Port Cisco Nexus 9372 B Eth1/1 10GbE NetApp controller 2 e0g Eth1/2 10GbE NetApp controller 2 e1b Eth1/3 10GbE NetApp controller 1 e0g Eth1/4 10GbE NetApp controller 1 e4b Eth1/5 10GbE NetApp controller 2 e0h 10 3D FlexPod: 3D VDI Visualization with TGX on

11 Eth1/6 10GbE NetApp controller 2 e4b Eth1/7 10GbE NetApp controller 1 e0h Eth1/8 10GbE NetApp controller 1 e1b Eth1/17 10GbE Cisco UCS fabric interconnect A Eth2/1 Eth1/18 10GbE Cisco UCS fabric interconnect A Eth2/2 Eth1/19 10GbE Cisco UCS fabric interconnect B Eth2/3 Eth1/20 10GbE Cisco UCS fabric interconnect B Eth2/4 Eth1/49 40GbE Cisco Nexus 9372 B Eth1/49 Eth1/50 40GbE Cisco Nexus 9372 B Eth1/50 MGMT0 GbE GbE management switch Any Table 3) NetApp controller 1 cabling information. Local Device Local Port Connection Remote Device Remote Port NetApp AFF8080 node 1 e0m 100MbE 100MbE management switch Any e0i GbE GbE management switch Any e0p GbE SAS shelves ACP port e0a 10GbE NetApp controller 2 e0a e0b 10GbE NetApp controller 2 e0b e0c 10GbE NetApp controller 2 e0c e0d 10GbE NetApp controller 2 e0d e0e 10GbE Cisco Nexus 9372 A Eth1/3 e0g 10GbE Cisco Nexus 9372 B Eth1/3 e4a 10GbE Cisco Nexus 9372 A Eth1/4 e4b 10GbE Cisco Nexus 9372 B Eth1/4 e0f 10GbE Cisco Nexus 9372 A Eth1/7 e0h 10GbE Cisco Nexus 9372 B Eth1/7 e1a 10GbE Cisco Nexus 9372 A Eth1/8 e1b 10GbE Cisco Nexus 9372 B Eth1/8 Note: e0m refers to the physical Ethernet port indicated by a wrench icon on the rear of the chassis. Table 4) NetApp controller 2 cabling information. Local Device Local Port Connection Remote Device Remote Port NetApp AFF8080 node 2 e0m 100MbE 100MbE management switch Any 11 3D FlexPod: 3D VDI Visualization with TGX on

12 e0i GbE GbE management switch Any e0p GbE SAS shelves ACP port e0a 10GbE NetApp controller 2 e0a e0b 10GbE NetApp controller 2 e0b e0c 10GbE NetApp controller 2 e0c e0d 10GbE NetApp controller 2 e0d e0e 10GbE Cisco Nexus 9372 A Eth1/1 e0g 10GbE Cisco Nexus 9372 B Eth1/1 e1a 10GbE Cisco Nexus 9372 A Eth1/2 e1b 10GbE Cisco Nexus 9372 B Eth1/2 e0f 10GbE Cisco Nexus 9372 A Eth1/5 e0h 10GbE Cisco Nexus 9372 B Eth1/5 e4a 10GbE Cisco Nexus 9372 A Eth1/6 e4b 10GbE Cisco Nexus 9372 B Eth1/6 Table 5) Cisco UCS fabric interconnect A cabling information. Local Device Local Port Connection Remote Device Remote Port Cisco UCS fabric interconnect A Eth2/1 10GbE Cisco Nexus 9372 A Eth1/17 Eth2/2 10GbE Cisco Nexus 9372 A Eth1/18 Eth2/3 10GbE Cisco Nexus 9372 B Eth1/19 Eth2/4 10GbE Cisco Nexus 9372 B Eth1/20 Eth1/1 10GbE Cisco UCS chassis 1 FEX A IOM 1/1 Eth1/2 10GbE Cisco UCS chassis 1 FEX A IOM 1/2 Eth1/3 10GbE Cisco UCS chassis 1 FEX A IOM 1/3 Eth1/4 10GbE Cisco UCS chassis 1 FEX A IOM 1/4 Eth1/5 10GbE Cisco UCS chassis 2 FEX A IOM 1/1 Eth1/6 10GbE Cisco UCS chassis 2 FEX A IOM 1/2 Eth1/7 10GbE Cisco UCS chassis 2 FEX A IOM 1/3 Eth1/8 10GbE Cisco UCS chassis 2 FEX A IOM 1/4 Eth1/9 10GbE Cisco UCS chassis 3 FEX A IOM 1/1 Eth1/10 10GbE Cisco UCS chassis 3 FEX A IOM 1/2 Eth1/11 10GbE Cisco UCS chassis 3 FEX A IOM 1/3 12 3D FlexPod: 3D VDI Visualization with TGX on

13 Eth1/12 10GbE Cisco UCS chassis 3 FEX A IOM 1/4 Eth1/13 10GbE Cisco UCS chassis 4 FEX A IOM 1/1 Eth1/14 10GbE Cisco UCS chassis 4 FEX A IOM 1/2 Eth1/15 10GbE Cisco UCS chassis 4 FEX A IOM 1/3 Eth1/16 10GbE Cisco UCS chassis 4 FEX A IOM 1/4 MGMT0 GbE GbE management switch Any L1 GbE Cisco UCS fabric interconnect B L1 L2 GbE Cisco UCS fabric interconnect B L2 Table 6) Cisco UCS fabric interconnect B cabling information. Local Device Local Port Connection Remote Device Remote Port Cisco UCS fabric interconnect B Eth2/1 10GbE Cisco Nexus 9372 A Eth1/17 Eth2/2 10GbE Cisco Nexus 9372 A Eth1/18 Eth2/3 10GbE Cisco Nexus 9372 B Eth1/19 Eth2/4 10GbE Cisco Nexus 9372 B Eth1/20 Eth1/1 10GbE Cisco UCS chassis 1 FEX B IOM 2/1 Eth1/2 10GbE Cisco UCS chassis 1 FEX B IOM 2/2 Eth1/3 10GbE Cisco UCS chassis 1 FEX B IOM 2/3 Eth1/4 10GbE Cisco UCS chassis 1 FEX B IOM 2/4 Eth1/5 10GbE Cisco UCS chassis 2 FEX B IOM 2/1 Eth1/6 10GbE Cisco UCS chassis 2 FEX B IOM 2/2 Eth1/7 10GbE Cisco UCS chassis 2 FEX B IOM 2/3 Eth1/8 10GbE Cisco UCS chassis 2 FEX B IOM 2/4 Eth1/9 10GbE Cisco UCS chassis 3 FEX B IOM 2/1 Eth1/10 10GbE Cisco UCS chassis 3 FEX B IOM 2/2 Eth1/11 10GbE Cisco UCS chassis 3 FEX B IOM 2/3 Eth1/12 10GbE Cisco UCS chassis 3 FEX B IOM 2/4 Eth1/13 10GbE Cisco UCS chassis 4 FEX B IOM 2/1 Eth1/14 10GbE Cisco UCS chassis 4 FEX B IOM 2/2 Eth1/15 10GbE Cisco UCS chassis 4 FEX B IOM 2/3 Eth1/16 10GbE Cisco UCS chassis 4 FEX B IOM 2/4 MGMT0 GbE GbE management switch Any 13 3D FlexPod: 3D VDI Visualization with TGX on

14 L1 GbE Cisco UCS fabric interconnect B L1 L2 GbE Cisco UCS fabric interconnect B L2 3.1 NVIDIA GPU Cards and NetApp FlexPod Installation and Configuration A sample installation and configuration of NVIDIA GPU cards with FlexPod Express and FlexPod Datacenter are described in the following links: TR-4563: How to Configure NVIDIA GPU M6 Cards with vsphere 6.0 in FlexPod Datacenter with Cisco B-Series Servers TR-4536: How to Configure NVIDIA GPU K1 and K2 Cards with vsphere 6.0 in FlexPod Express with Cisco C-Series Servers Since these two reference architectures were released, NVIDIA, Cisco, and NetApp have refreshed their equipment models. Therefore, NetApp recommends using C-Series 240M4/460M4 with NVIDIA M60 vgpu cards and A300 or A700 storage platforms for most high-end graphics rendering applications. Even though these links are based on other infrastructure, you can use them to understand the installation process with Cisco UCS C-Series servers or Cisco UCS B-Series servers and NetApp All Flash FAS storage. In addition, these links cover NetApp storage configuration and volume layouts. Note: 3.2 TGX NVIDIA provides different vgpu profiles depending on the type of graphics and the NVIDIA cards used. Contact Mechdyne to obtain the best possible vgpu profile for TGX and your vgpu application usage. For a full list of vgpu profiles for vsphere, see the NVIDIA GRID vgpu Deployment Guide for VMware. TGX Remote Desktop delivers graphics-intensive data from a remote location in high resolution without sacrificing image quality or user experience. The improved visual output and control give users the confidence to make critical decisions based on what they see on the desktop. In addition, TGX enables multiple collaborators to connect and control the session with the same level of performance as the local machine. 14 3D FlexPod: 3D VDI Visualization with TGX on

15 Figure 3) Dual 3840 x 1600 monitors displaying Schlumberger s Petrel software at 60 FPS using TGX with FlexPod. For more information, see video about 3D FlexPod. Download and Install TGX Software TGX installers are available from the Mechdyne Product Download Center. Contact Mechdyne to receive your login credentials and software licenses. TGX provides separate installers for the sender and the receiver. The sender is the remote workstation whose desktop and applications are shared by TGX with a receiver. The receiver is the local computer that displays and interacts with the remote desktop of the sender through TGX. Figure 4) Communication between sender and receiver TGX Windows Installation License Management: Sender Only The TGX sender requires a valid license to function. On Windows, install a license file (for example, TGX.lic) to C:\ProgramData\Mechdyne\licenses. To set an alternative license file location, set the LM_LICENSE_FILE or MECHDYNE_LICENSE_FILE environment variable. The environment variable 15 3D FlexPod: 3D VDI Visualization with TGX on

16 can provide a list of license files by using the ; separator or a list of servers (for example, port@hostname) by using the, separator. To set an environment variable on Windows, press the Windows key and type System to find the System control panel item. Select Advanced System Settings > Advanced tab > Environment Variables and add a new variable under System Variables. Connection Broker TGX can be independent of or integrated with connection brokers. The recommended solution is the Leostream Connection Broker because it provides a robust yet flexible solution for VDI environments. For more information, see the Leostream website TGX Windows Installation Windows System Requirements Operating System The TGX sender and receiver support Windows 7, 8, 8.1, 10, Server 2008R2, and Server Hardware and Drivers: TGX Windows Installation Table 7 lists hardware and drivers required for the TGX Windows installation. Table 7) Hardware and drivers for TGX Windows installation. TGX sender TGX receiver Graphics Card NVIDIA Quadro NVIDIA GRID (pass-through and vgpu) NVIDIA GTX Intel HD, Iris, Iris Pro NVIDIA Quadro NVIDIA GTX Intel HD, Iris, Iris Pro AMD (software decode) Driver NVIDIA R346 or later 354.xx recommended NVIDIA R346 or later 354.xx recommended TGX currently uses software decoding on AMD graphics cards. Hardware decoding is under development. Sender Installation For full sender installation instructions, see the following links: TGX User Guide Install Guide for Linux Install Guide for Mac Install Guide for Windows D FlexPod: 3D VDI Visualization with TGX on

17 Allow TGX to Overwrite the Existing Display Configuration TGX provides support for the sender desktop to be reconfigured to match that of the receiver. This option is enabled by default on the sender as part of the installation. To disable this capability, deselect this option in the TGX Sender configuration screen. NetApp recommends disabling this option on senders that are physically connected to complex display systems, especially those configured with NVIDIA Quadro Sync or Mosaic modes. Log Remote Users in Automatically After Authenticating TGX provides support to transfer the credentials of the user authenticated on the receiver to allow automatic login on the sender. This option is enabled by default on the sender as part of the installation. To disable this capability, deselect this option in the TGX Sender configuration screen. Port Configuration TGX reserves 16 ports for communication between the sender and receiver. The default starting port is However, this is configurable as part of the installation. Note that the default port must match between the sender and the receiver. Firewall Configuration By default, the installation process creates an exception to the standard Windows firewall rules to allow TGX connections. Uncheck this box if your IT security does not use the Windows firewall and have your IT group manually configure your firewall system to allow TGX connections. To install and configure TGX Sender, complete the following steps: 1. Install the TGX sender by using the setup wizard. Click Next. 2. Accept the EULA. 17 3D FlexPod: 3D VDI Visualization with TGX on

18 3. Select the destination location. 4. Allow TGX to overwrite the display configuration. 18 3D FlexPod: 3D VDI Visualization with TGX on

19 5. Select Log Remote Users in Automatically After Authenticating. 6. Confirm the port configuration. 19 3D FlexPod: 3D VDI Visualization with TGX on

20 7. Select Add Firewall Exceptions. 8. Select the Start Menu folder. 20 3D FlexPod: 3D VDI Visualization with TGX on

21 9. Click Install to initiate the installation process. 10. Complete installation. 21 3D FlexPod: 3D VDI Visualization with TGX on

22 Receiver Installation Port Configuration TGX reserves 16 ports for communication between the sender and the receiver. The default starting port is This is configurable as a part of the installation. Note that the default port must match between the sender and the receiver. Firewall Configuration By default, the installation process creates an exception to the standard Windows firewall rules to allow TGX connections. Uncheck this box if your IT security does not use the Windows firewall and have your IT group manually configure your firewall system to allow TGX connections. 1. Install the TGX receiver by using the setup wizard. Click Next. 22 3D FlexPod: 3D VDI Visualization with TGX on

23 2. Accept the EULA. 3. Select the destination location. 23 3D FlexPod: 3D VDI Visualization with TGX on

24 4. Select Add Firewall Exceptions. 5. Confirm the port configuration. 24 3D FlexPod: 3D VDI Visualization with TGX on

25 6. Select the Start Menu folder. 7. Select additional tasks. 25 3D FlexPod: 3D VDI Visualization with TGX on

26 8. Click Install to initiate the installation process. 9. Complete installation. 26 3D FlexPod: 3D VDI Visualization with TGX on

27 3.2.3 TGX Linux Installation Linux System Requirements Operating System The TGX sender and the TGX receiver support RHEL/CentOS 6 and 7, Ubuntu 16, and derivatives. The installer recognizes RHEL, CentOS, Scientific Linux, and Ubuntu explicitly and makes best guesses when configuring derivative distributions. Hardware and Drivers: TGX Linux Installation Table 8 lists the hardware and drivers required for TGX Linux installation. Table 8) Hardware and drivers for TGX Linux installation. TGX sender Graphics Card NVIDIA Quadro NVIDIA GRID (in pass-through mode) NVIDIA GTX Driver R346 and later 354.xx recommended TGX receiver OpenGL 2.1 support N/A The TGX sender requires an NVIDIA Quadro, GRID (in pass-through mode), or GTX graphics card with proprietary NVIDIA drivers installed. Open-source Nouveau drivers are not supported. The TGX receiver is currently limited to software decode on Linux, but requires a graphics card with OpenGL 2.1 support. Most graphics cards in the last decade support OpenGL D FlexPod: 3D VDI Visualization with TGX on

28 Install NVIDIA Proprietary Drivers TGX requires that the proprietary NVIDIA drivers for Linux are installed. For RHEL/CentOS, the drivers can be obtained from the NVIDIA website. For Ubuntu, the drivers can be installed from the Ubuntu repository that is provided within the Ubuntu operating system. The following are prerequisites are to install drivers for RHEL/CentOS: X is not currently running. Open-source Nouveau drivers are not active. The Linux kernel development package and GNU Compiler Collection are installed. Sender Installation Autoconfigure Xorg Configuration File The TGX installer checks for a functional NVIDIA driver and Xorg configuration. If the installer fails to find a valid Xorg configuration, it prompts you to generate one. Select Yes to generate a working Xorg configuration (/etc/x11/xorg.conf). The existing configuration is backed up. Allow TGX to Overwrite Existing Display Configuration TGX provides support for the sender desktop to be reconfigured to match that of the receiver. This option is enabled by default on the sender as part of the installation. To disable this capability, select No. NetApp recommends that this option be disabled on senders that are physically connected to complex display systems, especially those configured with NVIDIA Quadro Sync or Mosaic mode. Should TGX Start a New X Session? If no user is logged into an X session, the TGX sender, by default, starts a new X session for the connection rather than connecting to an existing (such as the display manager) session. This mode of operation is suitable for a server (either physical or VM) to which no displays are attached and X is not started on boot. For a desktop machine with connected displays, it might be desirable to use the existing X session. Select No to connect to the existing (display manager or user) X session rather than starting a new instance of X. On Ubuntu, this prompt is not displayed, and TGX always attempts to connect to the existing X session. Port Configuration TGX reserves 16 ports for communication between the sender and receiver. The default starting port is However, this is configurable as part of the installation. Note that the default port must match between the sender and the receiver. Firewall Configuration The TGX installer detects iptables (RHEL/CentOS 6), Firewalld (RHEL/CentOS 7), and UFW (Ubuntu 16) firewalls. It then installs the necessary exceptions to allow incoming connections on ports through (by default, or as set in the prior step). Additional network firewalls must be configured manually by the IT administrator. Start TGX Service? TGX fails to connect until a valid license is installed to /opt/mechdyne/licenses. See the section License Management: Sender Only for more information about license installation. Select Yes to start the TGX service. Otherwise, the TGX service can be manually started with the following commands: # systemctl start tgxserverd (RHEL/CentOS 7 and Ubuntu 16) # /etc/init.d/tgxserverd start (RHEL/CentOS 6) 28 3D FlexPod: 3D VDI Visualization with TGX on

29 1. Accept the EULA. 2. Confirm installation. 3. Accept the Xorg configuration. 4. Allow TGX to overwrite the display configuration. 5. Allow TGX to start a new X session. 29 3D FlexPod: 3D VDI Visualization with TGX on

30 6. Confirm the port configuration. 7. Configure the firewall (iptables, firewalld, or UFW). 8. Verify that a license is installed and start the TGX service. 9. Click OK when the installation is complete. Sender Installation Troubleshooting Disable X Start on Boot 1. Stop X. 30 3D FlexPod: 3D VDI Visualization with TGX on

31 # systemctl isolate multi-user.target (RHEL/CentOS 7 and Ubuntu 16) # telinit 3 (RHEL/CentOS 6) 2. For RHEL/CentOS 7, permanently disable X start on boot. # ln -sf /lib/systemd/system/multi-user.target /etc/systemd/system/default.target 3. For Ubuntu 16, permanently disable X start on boot. # ln -sf /lib/systemd/system/multi-user.target /lib/systemd/system/default.target 4. For RHEL/CentOS 6, edit /etc/inittab. Note: For a server or VM with no displays attached, NetApp recommends disabling X start on boot. Disable Nouveau ModeSet The NVIDIA drivers fail to install if the open-source Nouveau drivers are activated. To disable console mode setting by the Nouveau drivers, add nomodeset nouveau.modeset=0 to the kernel command line and reboot. For RHEL/CentOS 7, edit /etc/default/grub and add the arguments to GRUB_CMDLINE_LINUX. Then run grub2-mkconfig -o /boot/grub2/grub.cfg. For RHEL/CentOS 6, edit /boot/grub/grub.conf and add the arguments to the kernel command line. Xorg Configuration The TGX installer checks for a functional NVIDIA driver and Xorg configuration. If the installer fails to find a valid Xorg configuration, it prompts you to generate one. Answer yes to Xorg Configuration Not Found. Do You Want TGX to Configure X? to have the TGX installer generate a working Xorg configuration (/etc/x11/xorg.conf). The existing configuration is backed up to /etc/x11/xorg.conf.beforetgx. If the TGX installer does not prompt you to generate an Xorg configuration, remove or rename /etc/x11/xorg.conf and rerun the installer. Receiver Installation Port Configuration TGX reserves 16 ports for communication between the sender and the receiver. The default starting port is However, this is configurable as part of the installation. Note that the default port must match between the sender and the receiver. To configure ports on the receiver, complete the following steps: 1. Accept the EULA. 31 3D FlexPod: 3D VDI Visualization with TGX on

32 2. Confirm installation. 3. Confirm the port configuration. 4. Click OK. 32 3D FlexPod: 3D VDI Visualization with TGX on

33 3.2.4 TGX Mac Installation (Receiver Only) System Requirements The TGX Mac receiver supports OS X El Capitan Port Configuration TGX reserves 16 ports for communication between the sender and the receiver. The default starting port is However, this is configurable. The default port can be modified with the user interface or the command line or by manually editing the configuration file. All of these methods require elevated permissions. Note that the default port must match between the sender and the receiver. 1. On the TGX Launcher menu, select TGX > Preferences. Enter the desired port and click OK. 2. On the command line, navigate to the bundle and run the following command (as root): # TGX.app/Contents/MacOS/tgx_config_helper -p=<port>. 3. Edit /Library/Application Support/com.mechdyne.TGX/config.ini (requires elevated privileges), find the DefaultPort= line, and change the number to the desired port. 4. Install TGX. 5. Accept the EULA. 33 3D FlexPod: 3D VDI Visualization with TGX on

34 6. Select the installation type. 7. Click Close to complete the installation. 34 3D FlexPod: 3D VDI Visualization with TGX on

35 3.3 Documentation Links TGX User Guide Install Guide for Linux Install Guide for Mac Install Guide for Windows TGX FAQs TGX Troubleshoot Performance Optimization We have configured best-in-class algorithms to maximize the efficiency of frame buffer capture, encoding, decoding, mouse capture, and network management. We have leveraged hardware and software optimizations available on NVIDIA Kepler, Maxwell, and Pascal GPUs (Quadro and GTX) and on Intel GPUs. As a result, TGX delivers maximum frame rate with minimal latency and bandwidth consumption Latency Every network exhibits an inherent latency. The hardware on which TGX runs can also affect latency. TGX minimizes additional latency due to optimized encoding and decoding. The experience is as close to sitting at the remote computer as possible. The base latency is approximately the network latency plus D FlexPod: 3D VDI Visualization with TGX on

36 to 2 frame periods. For example, with 20ms of network latency and a frame rate of 48Hz, the latency should be approximately 40ms to 60ms. 3.5 Technical Features Encoding and Decoding TGX uses H.264 compression for all video encoding and AAC compression for all audio encoding. H.265 compression is under development. TGX detects the hardware configuration present for both the sender and receiver to make sure that the optimal path for encoding and decoding is taken and therefore maximum performance delivered. Bandwidth Consumption TGX uses variable bit-rate encoding to reduce the bandwidth requirement as much as possible. This means that, if there is no change in the image being sent, then the bandwidth is very low (generally < 1Mbps). For full-screen video or for applications for which much of the screen is changing, Table 9 shows expected bit rates. Table 9) Bandwidth consumption. Desktop Size Image Quality Average Bandwidth HD (1920x1080) at 60Hz Minimum 2.3Mbps HD (1920x1080) at 60Hz Default 35Mbps 4K (3840x2160) at 60Hz Minimum 3Mbps 4K (3840x2160) at 60Hz Default 45Mbps In general, larger displays, higher frame rates, and higher image quality require more bandwidth. In comparison with several competitive products, TGX uses 50% to 70% less bandwidth while delivering a higher image quality at faster frame rates IT Management and Security Authentication A TGX user must authenticate to a TGX sender by using a local account on the sender or with a domain account (Active Directory). For local accounts, when connecting with the TGX receiver, prefix the user name with.\ to indicate that the account is not on the domain. Network Security and Firewall TGX senders require open ports in the firewall for the receiver to connect. Mechdyne works with customers to make sure that all security requirements are met. TGX senders require ports through to be open for incoming TCP connections. Data Security TGX encrypts all network traffic with Secure Sockets Layer (SSL) encryption. This encryption protects the privacy of display, keyboard, mouse, and audio information between the sender and the receiver. Transient (in-memory) copies of credentials (domain, user name, and password) are hashed with 256-bit Advanced Encryption Standard (AES-256) in all TGX applications. Any interprocess communication between TGX applications over the network or on the same computer is also 256-bit AES hashed. By 36 3D FlexPod: 3D VDI Visualization with TGX on

37 default, all TGX network traffic uses SSL (4096-bit RSA security key) for encryption. There are no differences in encryption between Windows, Linux, and Mac. When TGX is used with a broker, the connection file uses 4096-bit RSA public/private key encryption for the password field (domain and user name are plain text). For the best security, no one should keep a TGX file with any form of password (plain or encrypted) on disk. 4 Conclusion 4.1 Use Cases This reference architecture is intended for high-end graphics users and graphics applications for which 3D graphics rendering is a requirement. For example, oil and gas researchers might want to visualize ground formations, or aerospace engineers might want to visualize air velocity on an aircraft wing. DNA sequences can be graphed out in biomedicine, or computation flow dynamics can be analyzed in many industries. Universities can teach students how to use 3D application rendering. These are just a small sampling of the types of industries and applications in which this reference architecture can be used. The following list is a sample of the industries and use cases for which this reference architecture is intended: Aerospace applications Oil and gas applications Biomedical research applications Municipal water flow and distribution research Any computational flow dynamics applications Higher education and universities 4.2 Sizing Sizing the number of VDI sessions per compute node, per GPU card, and per storage array depends on several performance metrics. To size storage, contact your NetApp account team. They can estimate the number of IOPS and the size of total files located on the NetApp storage. The account team uses the NetApp System Performance Modeler (SPM) sizing tool to determine the model and capacity of the NetApp storage. To size NVIDIA GPU cards and Cisco UCS compute nodes, contact your account representative at Mechdyne. Mechdyne can determine the number of users who can be accommodated with each compute node or GPU card. The sizing for compute nodes or GPU cards is determined by the type of application to be used, the size of video input processing files, and the desired resolution rate of the video (4K and other options). Table 10 provides a sizing guideline, but your design must be validated for your use case by your NetApp and Mechdyne account teams. Table 10) Sizing information. NVIDIA Card per Compute Node Low-End Graphics Users (PowerPoint, Office, and So On) Midlevel Graphics Users (Medium Resolution) High-End Graphics Users (4K Resolution) NVIDIA M60 card (with NVENC chip) NVIDIA M10 card (with NVEC chip) 16 users 8 users 4 users 8 users 4 users 2 users 37 3D FlexPod: 3D VDI Visualization with TGX on

38 NVIDIA M6 card (no NVEC chip) 16 users 8 users Not available Note: Sizing can vary substantially between graphics applications. To obtain sizing for your video graphics users needs, contact your NetApp and Mechdyne sales teams. 4.3 Summary The combination of Mechdyne TGX software with Cisco UCS Manager, Cisco UCS B200 M4 rack servers, NVIDIA GRID M60 cards, and VMware vsphere ESXi 6.0 provides a high-performance platform for virtualizing graphics-intensive applications. Most companies are looking at flash for their graphics workstations or VDI environments. However, you must design a solution that includes applications and data together on flash to provide the end user with the best experience. NetApp All Flash FAS (AFF) provides high performance for the processing of graphics files on CIFS shares and the VDI environment for shared storage. The combination of both Cisco and NetApp in a FlexPod Datacenter converged infrastructure provides superior performance for GPU users. With the guidance in this document, our customers and partners are ready to host a growing list of graphics applications. The following key findings were observed during the reference architecture configuration: Use the software installation prescribed in this document for 3D FlexPod configuration. Cisco UCS servers provide superior compute node performance. Cisco Nexus switches provide superior network performance. Use NetApp storage for the large graphics files. The use of SSDs is extremely beneficial, with less than 1ms access. NetApp support for multiple protocols removes the need for multiple storage systems. NetApp CIFS supports profiles, home directories, and the graphics files to be processed. NetApp CIFS provides very high performance for processing large video graphics data workloads. TGX software provides the graphics visualization rendering and performance required for high-end video clusters. References This section lists the resources we used to build out the environment. 3D FlexPod TR-4563: How to Configure NVIDIA GPU M6 Cards with vsphere 6.0 in FlexPod Datacenter with Cisco B-Series Servers TR-4536: How to Configure NVIDIA GPU K1 and K2 Cards with vsphere 6.0 in FlexPod Express with Cisco C-Series Servers Cisco UCS C-Series Rack Servers Cisco UCS B-Series Servers Cisco UCS Manager Configuration Guides D FlexPod: 3D VDI Visualization with TGX on

39 Related FlexPod with Cisco UCS Guides Deployment Guide for FlexPod with VMware vsphere 6.0 and NetApp AFF 8000 Series and Cisco Nexus 9000 Series Switches for Top of Rack FlexPod Datacenter with VMware vsphere 6.0 Design Guide Site Requirements Guide NVIDIA NVIDIA GRID Resources NVIDIA GRID K1 AND K2 Graphics-Accelerated Virtual Desktops and Applications Tesla M6 Product Brief NetApp Storage and Storage Interconnect Switches Related FlexPod with Cisco UCS Guides Deployment Guide for FlexPod with VMware vsphere 6.0 and NetApp AFF 8000 Series and Cisco Nexus 9000 Series Switches for Top of Rack FlexPod Datacenter with VMware vsphere 6.0 Design Guide Site Requirements Guide Clustered Data ONTAP High-Availability Configuration Guide Clustered Data ONTAP Network Management Guide Clustered Data ONTAP Software Setup Guide Storage Subsystem Configuration Guide Advance Drive Partitioning FAQ FlexPod Datacenter with VMware vsphere 6.0 Design Guide FlexPod Datacenter with VMware vsphere Clustered Data ONTAP NFS Best Practice and Implementation Guide NetApp Data Compression and Deduplication Data ONTAP and Later TGX Remote Desktop TGX User Guide Install Guide for Linux Install Guide for Mac D FlexPod: 3D VDI Visualization with TGX on

40 Install Guide for Windows TGX FAQs TGX Troubleshooting D FlexPod: 3D VDI Visualization with TGX on

41 Recognition Authors Chris Rodriguez, Senior Technical Marketing Engineer, NetApp, Inc. Chris (C-Rod) is a senior technical marketing engineer (TME) at NetApp. Chris has been involved with VDI since the late 1990s. Prior to Chris s TME role, he worked as a NetApp Professional Services consultant and implemented NetApp storage in VDI environments. He has 15 years of enterprise storage experience and has architected, designed, and implemented storage for many enterprise customers. Currently, Chris works for the Converged Infrastructure team at NetApp, where he develops reference architectures with VDI on NetApp storage and FlexPod. Mark Hirst, Software Sales and Marketing Strategist, Mechdyne, Inc. Mark Hirst is a senior sales and marketing strategist at Mechdyne. After 10 years of global account management with enterprise customers, Mark transitioned to lead the sales and marketing efforts of Mechdyne s software division. Mark s role encompasses the full range of Mechdyne s advanced visualization and virtual reality products. His recent focus has been on TGX, Mechdyne s advanced remote desktop solution. Acknowledgements The authors would like to thank the following people: Josh Devan, Mechdyne technical writer and TGX team member Chris Gebhardt, NetApp end-user computing product manager and principal technical marketing engineer Cathy Joyner, Mechdyne technical writer and TGX team member Troy Magnum, NetApp cloud convergence infrastructure (CCI) manager Joe Moder, Mechdyne technical writer and TGX team member Osama Qazi, consulting systems engineer Kenneth Westerby, NetApp 3D FlexPod evangelist 41 3D FlexPod: 3D VDI Visualization with TGX on

42 Refer to the Interoperability Matrix Tool (IMT) on the NetApp Support site to validate that the exact product and feature versions described in this document are supported for your specific environment. The NetApp IMT defines the product components and versions that can be used to construct configurations that are supported by NetApp. Specific results depend on each customer s installation in accordance with published specifications. Copyright Information Copyright NetApp, Inc. All rights reserved. Printed in the U.S. No part of this document covered by copyright may be reproduced in any form or by any means graphic, electronic, or mechanical, including photocopying, recording, taping, or storage in an electronic retrieval system without prior written permission of the copyright owner. Software derived from copyrighted NetApp material is subject to the following license and disclaimer: THIS SOFTWARE IS PROVIDED BY NETAPP AS IS AND WITHOUT ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE, WHICH ARE HEREBY DISCLAIMED. IN NO EVENT SHALL NETAPP BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. NetApp reserves the right to change any products described herein at any time, and without notice. NetApp assumes no responsibility or liability arising from the use of products described herein, except as expressly agreed to in writing by NetApp. The use or purchase of this product does not convey a license under any patent rights, trademark rights, or any other intellectual property rights of NetApp. The product described in this manual may be protected by one or more U.S. patents, foreign patents, or pending applications. RESTRICTED RIGHTS LEGEND: Use, duplication, or disclosure by the government is subject to restrictions as set forth in subparagraph (c)(1)(ii) of the Rights in Technical Data and Computer Software clause at DFARS (October 1988) and FAR (June 1987). Trademark Information NETAPP, the NETAPP logo, and the marks listed at are trademarks of NetApp, Inc. Other company and product names may be trademarks of their respective owners. 42 3D FlexPod: 3D VDI Visualization with TGX on