Before the release call you can save about 45 minutes of time by performing following steps:

  • Download SAFR Platform CUDA 11 Version from on all three VM instances
  • Run the installer on all three VM instances (DO NOT Sign in to ANY of them)


Release Prep Checklist

[ ] Machines or VMs are Provisioned with required operating system

[ ] Nvidia GPUs are installed (if using VMs Nvidia VGPU licenses installed)

[ ] nvidia-smi utility reporting GPUs are visible to OS (validate decode is working and putting load on GPUs)

[ ] (Optional) and safr.real.com443 is whitelisted (this will make it much easier to install, license and upgrade server)

[ ] SAFR Ports on internal network are opened

[ ] Windows machine available on the network to install and test SAFR Desktop 

[ ] At least one IP camera is available and accessible on the network for testing


Ideally we have the following expertise available on the release call:

  • Administrative control over Servers
  • Expertise in the Operating system in use
  • If VM, Expertise in the VM technology
  • Expertise in Nvidia drivers for the OS in question (especially for Linux)
  • If VM, expertise in Nvidia Virtual GPU Software, drivers and licensing
  • Network engineer who can investigate or make change to networks/switches
  • If Load balancing, someone with access and expertise in load balancer