IDLive Face Server¶

IDLive Face Server bundles IDLive Face as a web application you can deploy in your data center or private cloud. It is distributed as a container image that can be used with Docker or Kubernetes.

Hardware requirements¶

For the CPU version you need the CPU that supports at least AVX2 instruction set. More cores and higher clock rates will yield better processing times. We recommend using as many cores as you possibly can, but note that NUMA configurations are not supported. Make sure that you use the single-socket system. The memory usage depends on the number of CPU cores and the size of a workload. We recommend to start from 4 Gb and increase it if needed.

For the GPU version you need the Nvidia GPU with the compute capability from 7.0 to 8.9. You also need a reasonable CPU, to make sure that workload for the GPU is generated fast enough. We recommend to have 2 cores if you plan to process a single image only, and up to 8 if you plan to process several images at once. As with the CPU version the memory requirements depend on your workload, but for the GPU version you should start from 8 Gb. Your GPU also needs to have at least 8 Gb of memory.

If you plan to use a cloud provider, we recommend compute-optimized instances, like AWS C5 and AWS g5 or Azure Fsv2.

On AWS we recommend using c5.4xlarge for perfomance or c5.12xlarge for workload. If running the server on a GPU machine, g5.2xlarge is a good choice.