In production ,is running multi triton-inference-server on 1 GPU a good idea?

[oymy]

what is the pros&cons of running multi triton-inference-server on 1 GPU,
would it causing low gpu utilization ?
any article discussing about the pros & cons.

thanks.