Now you may remember my blog post from February, when I uncovered my identity as a XenAgent in the XenServer Master Class. But I also work as an HDX Agent. Am I a double agent? Not really, Citrix XenDesktop, XenServer & HDX 3D Pro have a long history of working together, since we introduced GPU Passthrough with XenServer 5.6 FP1.
If you never attended a Master Class but you are interested in a Citrix technology you really should go check out one of our Master Classes. It is a 2+ hours Webinar hosted by internal technology specialists like Support, Readiness or Product Management. We will show live demos, hints and best practices. And also we have regular Guest Speakers that provide add-on Products or services around Citrix Products.
A Bonus of these Master Classes is that you can ask questions during the Master Class, which will be answered by our Field Specialists (like me).
Now for some legal regulations, we cannot share the full Q&A but I going to share 5 of the common questions we received.
Question: What servers can be used with this GRID technology?
Answer: We provide a full list of certified servers for the vGPU technology at our XenServer Hardware Compatibility List: http://hcl.xensource.com/vGPUDeviceList.aspx
Question: Is there any difference between ICA and HDX?
Answer: HDX is the umbrella term for the enhancements on top of ICA, that enable user experience optimization and additional functionality such as printing, 3D Space-mouse support, USB support, WAN, etc. For more information you may want to check: https://www.citrix.de/content/dam/citrix/en_us/documents/products-solutions/citrix-hdx-technologies.pdf
Question: How are resources shared on K2 or is it evenly split?
Answer: You can split eachGPU on the K1 and K2 cards, into evenly sized vGPUs. These are called vGPU profiles. You can learn more about the different profiles and there targeted users: http://www.nvidia.com/object/virtual-gpus.htmlThis means on a K2 card with 2GPUs, you can allocate one GPU to K260Q profiles for high end-CAD designers and the other to K200 profiles for less graphically demanding windows desktops delivering applications such as Microsoft Office.
If the GPU is under-utilised, the intelligent resource sharing will allow the GPUs in use to access the spare GPU cores and you should not be surprised to see a GPU consume >50% of the resource in this scenario even if the GPU is allocated to be shared.
Question: If the end user is working with an ray-tracing app that relies GPU rendering power, would it be possible to assign more then one GPU to a VM?
Answer: Officially this is not supported, but it works. Please test it how it works for you. Please just follow the Guidance from Rachel‘s Blog: /blogs/2014/07/25/assigning-multiple-gpus-to-a-vm-with-citrix-xenserver/
Partner Demos and Case Studies
During the class we saw some great case studies and demos from partners such as IMSCAD (who specialise in Autodesk virtualisation) and Lakeside whose SysTrack product can be used for general but also GPU monitoring and also auto-capacity planning.
We also saw a preview of Cisco’s new white paper, „Cisco UCS C240-M3 Rack Server with NVIDIA GRID GPU cards on Citrix XenServer 6.2 and XenDesktop 7.5“, now available – a great read which you can access, here. It’s great to see Cisco investing in these technologies and my colleagues have blogged about their great servers designed for these technologies. Cisco have done a lot of work with these technologies and there is some nice material available on Youtube as well as the traditional whitepapers.
Want to ask more questions?
If you can’t wait until our next class, we have lots of blogs answering common questions and the authors are usually very responsive in answering relevant questions and comments, just search for the technology or applications you are interested in!