Two years ago I wrote an article about running Teradici’s Cloud Access Software on Microsoft Azure with the release of N-series instances (https://msandbu.org/test-run-of-teradici-cloud-access-software-on-azure-n-series/) now the issue with the N-series is that they don’t have the ability to provide SSD functionality, which was one of my biggest concerns. Now in a later stage, Microsoft has now released more GPU nodes which provides more CPU, more Memory and more GPU cards with SSD support. Still, Microsoft lacks a lot of flexibility when it comes to IaaS and hence the subject on this post.
I’ve also been writing a bit about Google Cloud Platform lately and how its IaaS model works (Combining customizable virtual machines where you can define which CPU architecture, memory and disk type when setting up virtual machines. Also, Google has now gotten a quite extensive support when it comes to GPU offerings now.
Should also be noted that Google now provides sustained use discount on GPU cards as well –> https://cloud.google.com/compute/docs/sustained-use-discounts#sud_gpus
Also, they provide some limitations when it comes to which type of regions and zones that can be used to deploy an instance on Google Cloud –> https://cloud.google.com/compute/docs/gpus/#gpu-virtual-workstations
But as of now, running instances on Google Cloud with GPU is currently supported with VMware Horizon and Teradici Cloud Access. It should also be noted Citrix is also working on GCP support and as of now has Power Management capabilities with MCS support coming, but GPU support is still not available.
So what about the NVIDIA Grid license? Spinning up a predefined instance on GCP with GPU you will automatically get billed for that license as part of the monthly cost, and the
It is also important to note that when setting up a GPU instance on GCP you cannot live-migrate that VM when it comes to host maintenance using the built-in functionality in GCP. So important to remember that these maintenance events typically occur once per week, but can occur more frequently when necessary, so whatever application you have on GCP you need to ensure that it can restart on a schedule.
You can query the metadata information to get info when your VM is scheduled to reboot from with the VM host –>
curl http://metadata.google.internal/computeMetadata/v1/instance/maintenance–event –H “Metadata-Flavor: Google”
Setting up Teradici Cloud Access from Google can be done through the Deployment Manager or the marketplace.
After the deployment is done you essentially just install the agent. All firewall rules are automatically configured as part of the deployment template. The agent can be found
here –> https://docs.teradici.com/find/subscription/product/cloud-access-software/component
As part of the deployment, it also installs the NVIDIA Grid driver, as part of the template. 
So setting up a Teradici Solution on GCP is pretty straightforward and combined with the flexibility that Google provides with virtual machines it makes it a pretty compelling offering for power users that need single instance VDI or Session hosts powered by GPU.