LiveKit distributes calls among multiple worker instances using a built-in load balancing system: • Round-Robin Distribution: Jobs are distributed to workers in a round-robin manner to ensure even distribution. • Single-Assignment Principle: Each job is assigned to only one worker at a time. • Timeout Handling: If a worker doesn’t accept a job within a set time, the job is redirected to another available worker. • Geographic Affinity: LiveKit Cloud prioritizes workers closest to the user to reduce latency. • Worker Availability: Workers’ availability is determined by their CPU usage and a load threshold (default is 0.75), preventing overloaded workers from receiving new jobs. This ensures efficient and balanced call distribution across all available workers, maintaining optimal performance and minimizing user latency. Sources: Deploying to production | LiveKit Docs | Self-hosting the Egress Service | LiveKit Docs | Deployment and scaling | LiveKit Docs