Network Load Balancers
Network Load Balancers (NLB) control the routing of WebRTC traffic between the external client and the instance of the streaming Omniverse Kit Application running within the cluster.
You can configure the number of streams an NLB can manage, allowing for a variety of deployment options.
When deployed to AWS or Azure, the CSP native NLB is used. When deploying on-prem, you must decide which NLB best fits your needs. NVIDIA uses MetalLB for internal development and testing.
TLS Encryption
Omniverse Application Streaming API supports two methods of TLS encryption of the WebRTC TCP signaling channel; termination at the NLB or termination at the Pod through Envoy Proxy. For simplicity in management and deployment, the preferred method is to enable TLS termination at the NLB by attaching a valid TLS certificate.
Provisioning NLBs
You are responsible for provisioning the NLBs available to be used by the Omniverse Application Streaming API instance.
AWS NLB Manager (optional)
In AWS, Listeners
and Target Groups
must be created and properly configured to enable routing to Kit Applications. The Omniverse Application Streaming API contains an optional service, AWS NLB Manager
, that automates the creation and re-binding of these components as streams are initiated and terminated.
Tags are used to identify which NLBs in the cluster are available for WebRTC routing, allowing the AWS NLB Manager
to dynamically manage the appropriate resources.