Earlier this year, NVIDIA announced its’ new GRID technology. I’ve been hearing all these questions — “what’s all the hype?” “what does this mean for my environment?” “will this benefit me?”
I wasn’t too sure about any of it myself. The product overviews on their website are vague at best, simply spitting out information about watt per gigaflop. I’m fortunate enough to be at the NVIDIA GPU Tech Conference in San Jose, California. Over the past few days I’ve gained a lot of insight as to how the GRID technology works and how hugely beneficial it is going to be for power users in your virtualized environment.
So, is it worth all of the hype? In my opinion, absolutely. The stuff that these new GRID cards can do is absolutely incredible. So let’s dive right in and take a look at all of this stuff.
The problem with VDI is that it makes users take a step back. End users are used to having a full desktop computing experience, when you give them VDI, they feel like they’re losing a lot. They no longer have a fully dedicated piece of hardware for themselves. No big desktop with a quad core CPU at their disposal, no more dedicated graphics card, and that scares them. When you take their dektop and move it into the datacenter, they’re unhappy. End users just want it to work, and VDI has struggled in this department.
Why you ask? Think about it. Instead of a dedicated piece of hardware on their desk, you’re forcing them to share resources with others. You take, say, a 12 core processor and spread 24 users out across it. No longer do they have their full quad core processor at their disposal, they’re sharing a full core with someone else, essentially having half a core. Okay, big deal, they’re using word docs and browsing the internet. What many people fail to realize though is the fact that you’ve takin their GPU away too! Then how do we get graphics? Well, that half a core they’re using for processing is also being used to process and display the graphics. As you can see, we’ve taken a lot away from what users are used to having — so their experience suffers as we move them to VDI.
I think this is where NVIDIA GRID steps into the limelight. While it doesn’t solve the problem completely, it absolutely helps us IT pros out in delivering a better VDI experience to our users.
So, what do you need to bring GRID to your environment?
- A GRID enabled hypervisor & GRID GPU
- NVIDIA Driver
- NVIDIA GRID enabled virtual desktop
Let’s break these down a little bit. A GRID enabled hypervisor sounds scary. It’s not. Many products will work with the GRID technology, VMware’s ESXI, Citrix XEN, etc. You can check out the GRID webpage at the NVIDIA webpage if you would like to look at a compatibility matrix.
NVIDIA driver?! What’s this all about? This is pretty simple too. It is just a driver from NVIDIA that is injected into your hypervisor that allows for your VMs to pass through graphics processing to the dedicated GPU in your host. I can be direct passthrough 1:1 or the VMs on that host can share the graphics compute of the video card that you have installed in your host. Mind you, if a VM is vMotioned off this host onto another without a graphics card, that VM won’t be able to take advantage of said graphics compute anymore.
A NVIDIA GRID enable virtual desktop is pretty simple too. Usually just another driver that you have to install on your gold image. In the case of VMware, its an extra option you have to enable in the VMware View agent on your gold image.
Having gotten my hands on this technology, I can attest to the big boost in graphics power of the virtual desktop when a product like NVIDIA GRID is installed in your environment. I can say the highest amount of overhead on a physical host is the graphics processing. If you can offload this on a graphics card dedicated for graphics processing, not only can your density increase (in most cases), but your overall end user experience will improve.
Questions? Feel free to post in the comments section below.
Next order of business — I’ll define my setup, configurations, etc. in my VMware View and NVIDIA GRID environment.
Then, we’ll start a tutorial series on how to get it all setup with best practices and working in your environment!