NVIDIA GRID vGPU Technology on Citrix
This effort was started in 2006 with the Citrix/NVIDIA collaboration effort with Boeing for the Dreamliner 787. Boeing needed to have engineers from around the globe able to remotely collaborate on this project. The application they needed was Dassault’s CATIA 3D modeler. The NVIDIA GRID™ K1 and K2 vGPU cards are used along with XenDesktop 7 in XenApp or VDI mode. K2 cards are MUCH more powerful. Here are the main drivers for this need:
- A global talent base of engineers requires a good remote access experience for 3D modeling.
- Security of Intellectual Property (IP) becomes a concern as a result. Without a good WAN accessible application virtualization capability, companies are forced to transfer highly secret CAD files to remote engineers for local processing. This exposes the companies to risk of piracy. This solution allows centralization and protection of all IP.
- Improves operating efficiency and reduces costs. Another factor of using remote engineers without this solution is the time it takes to transfer models to them to work on them. This can take hours. Using this solution enables follow the sun development since the transfer problem is eliminated.
- Enables work from home and DR for obvious reasons.
- Need to Leverage mobile devices to show a drawing at a customer site, or look at a drawing on the shop floor. Clients are available for iOS and Android, as well as Windows, MAC and Linux.
The market for this type of solution is broken down as follows:
- Professional Users. Full-time designers, radiology technicians, etc. that need a dedicated GPU and 3D mouse support. Not likely candidates for this solution.
- Power Users. A large group of users that need to VIEW or EDIT large 3D models and are excellent candidates for GPU sharing.
- Knowledge Workers. These users are great candidates for highly shared GPUs.
This capability is called HDX 3D Pro and is a feature of XenDesktop 7 ENT and PLAT. It may only be run on certain qualified servers that can support these NVDIA cards. For us, it is the Dell R720. It is purchased as a package from Dell… server, NVIDIA card, and all, and that is the only way it is available.
There are 2 types of compression that make this possible:
- H.264 Compression is “Deep” compression that allows excellent performance, event at WAN speeds (as low as 1.5MBPS).
- Lossless (Picture Perfect) Compression for all those apps (think medical imaging) that MUST have perfect resolution.
Here is a description of how the product works in each environment.
HDX 3D Pro on XD 7 RDS (XenApp) –Available NOW (Up to 30 people sharing 1 GPU)
HDX 3D Pro on XD 7 VDI –Available NOW with GPU sharing (Up to 8 people sharing 1 GPU)
- GPU assist on Direct 3D, OpenGL , CUDA and OpenCL via GRID™ vGPU Cards (up to 8 per GPU)
- Support for 3D mouse
- Published app or published desktop
- H.264 and Lossless compression
- Support for all clients (Windows, Mac, Linux, iOS and Android)
- WINDOWs only apps, no Linux (yet)
- WINDOWs versions of apps are lower cost than Linux versions
- This solution suitable for Power users
- Popular Apps available for this platform are:
- PTC Creo
- AutoDesk
- Bentley
- BS Solidworks
- Siemens
- Adobe
New features of XD7 with NVIDIA over earlier versions of XD and NVIDIA cards are:
- No “API Hooking” is faster and leads to more compatible applications
- Self-tuning CODEC (adjusts to fluctuating bandwidth)
- Now available on XenApp
- 70% higher Frame Rates than with XD 5.6
- Auto screen resolution
- H.264 and lossless compression
- HDX monitor reports on the status and performance of each session
- Quad monitor support
- Runs on XenServer 6.1 or vSphere 5.1. No Hyper-v since they support Remote/FX.
- Supports “HDX Ready PREMIUM” and HDX SoC (CODEC support directly on a chip)
- Use of CloudBridge (NetScaler) gains up to 3-to-1 additional compression.