WebRTC at Scale: Transitioning from Jitsi to LiveKit on Kubernetes
Real-time video communication today relies almost exclusively on WebRTC. However, WebRTC is not a …

Imagine your company is an exhibitor at the year’s most important trade show. The booth is packed, your sales team is highly motivated, and at every terminal, a potential major client is waiting for a live presentation. At this moment, the infrastructure is the weakest link in the chain.
In traditional environments based on a few fixed servers, panic often ensues under this load: Performance drops, instances freeze, and new demos can no longer be launched. What was planned as a marketing highlight turns into a technical debacle. A modern, Cloud-native infrastructure turns this stress scenario into routine.
Why do conventional systems reach their limits with high concurrency?
With a platform based on Managed Kubernetes, you face the trade show rush with technological superiority.
Instead of hoping one server will suffice, the platform uses “Elasticity.” When the system detects that the load from many parallel demos is increasing, the cluster automatically books new computing capacity from the cloud in the background. Once the trade show is over, the system shrinks back to its base size. You only pay for the peak load when you actually need it.
By using namespaces and so-called Resource Quotas, a “safety container” is defined for each demo.
Since each demo is launched as a lightweight Container, you can start 40 instances almost simultaneously. While a conventional system would buckle under the tenth parallel installation, Kubernetes distributes the tasks efficiently.
A resilient infrastructure changes your team’s presence on-site:
Peak loads are the ultimate test for your IT strategy. Those who rely on modern platform principles turn the technical challenge of a trade show into a competitive advantage. Stability with high concurrency is no accident but the result of an architecture based on isolation and automated scaling.
No. That’s the beauty of GitOps and Kubernetes: The environment is the same as in everyday use. The only difference is that the cluster starts more nodes (servers) in the background to handle the mass of identical workflows.
The infrastructure runs stably in the cloud. If the on-site Wi-Fi fluctuates, the issue lies with the connection, not the server. We often recommend local 5G routers as a backup since the server-side performance is guaranteed by the platform.
Yes. Through the pipeline, you can pre-warm 20 instances with a command before the trade show starts, so they are ready to go when the first visitors arrive.
Once the demo instances are deleted (or reach their expiration date), the cluster autoscaler recognizes that the additional computing power is no longer needed and scales down the cloud instances. Costs immediately return to normal levels.
Real-time video communication today relies almost exclusively on WebRTC. However, WebRTC is not a …
Nothing is more frustrating for an operations team than a 3 AM alarm that turns out to be a …
When planning cross-site infrastructure, architects often face a fundamental decision: Do we …