Case Study
Case Study
Enhancing Application performance with KMS Monitor
Set up a Meeting
Kurento Media Server (KMS), a top-tier WebRTC media server in the MCU category, powers real-time multimedia applications. A client in the insurance sector faced a significant challenge with their conferencing application, which struggled to host multiple meetings on the media server and required frequent restarts, posing a major risk to their production system.
SpringCT investigated the system and identified the root cause: improper coding that prevented media pipelines—resources on the media server—from being released under certain conditions, leading to a gradual increase in server load. To resolve this, SpringCT developed KMS Monitor tool that can monitor media pipelines and endpoints present challenges, particularly with issues like lingering zombie endpoints after meetings.
Product Features
The KMS Monitor is a comprehensive monitoring tool
designed for Kurento Media Server. Its key features
include:
Pipeline Visualization
Real-time graphical representation of media pipelines
and connections between endpoints.
Media Flow Monitoring
Visual indicators (green for healthy, red for issues)
to show media flow directions and detect transmission
problems.
Detailed Stats Analysis
Key metrics such as bitrate, packet loss, and latency
for each connection.
WebRTC Negotiation Inspection
Displays offer and answer exchanges for effective
debugging.
PlayerEndpoint Attachment
Allows direct playback of media streams for inspection
and validation.
Key Technical Achievements
Zombie Pipelines and Endpoints
Zombie WebRTC endpoints and pipelines would
occasionally persist even after conferences ended,
consuming server resources unnecessarily and impacting
new sessions.
Lack of Real-Time Feedback
Without a monitoring solution, identifying and
diagnosing issues like packet loss, latency, or ICE
candidate failures required significant manual effort.
Complex Debugging
Troubleshooting WebRTC sessions without clear
visibility into offer/answer exchanges or endpoint
stats slowed resolution times.
Technologies Used
- Kurento API: Kurento Client JS APIs for monitoring pipeline components and Stats API for real-time statistics on media flows and connections.
- WebSocket Communication: Enables low-latency updates between KMS and the monitoring dashboard.
- Visualization Libraries: Tools like D3.js to dynamically render the pipeline connections and media flows.
- Backend Infrastructure: Node.js and Express for handling communication between the KMS Monitor and Kurento Media Server.
- UI Frameworks: Modern web technologies like React.js for an interactive and responsive user interface.
Results
- Detection of Zombie Endpoints: The monitor identified zombie pipelines and endpoints post-conference, enabling cleanup and freeing up server resources, which also helps finding defects in KMS pipeline architecture on the KMS App Server.
- Improved Debugging: With real-time stats and WebRTC negotiation insights, developers reduced troubleshooting times by 50%.
- Enhanced Media Quality: Continuous monitoring allowed for proactive identification of packet loss and bitrate issues, ensuring smoother media delivery.
- Operational Efficiency: Media operations teams gained better visibility into active sessions, reducing downtime and enhancing user experience.
Conclusion
- The KMS Monitor has proven to be an indispensable tool for managing Kurento Media Server instances.
- By detecting and addressing issues like zombie endpoints and providing real-time feedback on media flows, SpringCT could significantly improved system performance and reliability by fixing the code.
