Load Balancer Distributes Media Agents in Conference Systems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Cloud-based conference systems face challenges in assigning the best media agents to callers in highly distributed systems, managing available capacity, and rapidly recovering from failures due to outdated agent location databases and network separation delays, which leads to quality degradation and increased costs.
Innovation Solution
The solution involves delegating media agent discovery and configuration to media-plane signaling, using techniques like STUN URLs and load balancers to dynamically assign available media agents based on proximity and capacity, and implementing failover mechanisms for rapid recovery.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If cloud-based conference systems use centralized agent databases for media agent assignment, then system management is simplified, but agent location data becomes outdated and assignment accuracy deteriorates due to network separation delays
Solution Approach 1:
The system performs preliminary actions by having media agents proactively publish their availability and location information to the load balancer before being requested. This ensures the load balancer has up-to-date information without needing to query centralized databases, resolving the contradiction between simplified management and accurate real-time data.
Solution Approach 2:
The load balancer acts as an intermediary between callers and media agents, maintaining a local cache of agent availability information. This intermediary role allows the system to simplify centralized management while providing accurate real-time agent location data through the load balancer's local information store, eliminating network separation delays.
2Reliability
If cloud-based systems route all media traffic through centralized data centers, then quality of service is maintained, but wide area network bandwidth costs and latency increase
Solution Approach 1:
The system applies local quality by enabling media agents distributed in different geographic locations to handle media traffic locally for nearby callers. The load balancer directs callers to the nearest available media agent, allowing high-quality local media processing while minimizing wide area network bandwidth consumption for long-distance traffic.
Solution Approach 2:
The system transitions from a single centralized dimension to a multi-dimensional distributed architecture. Media agents are distributed across multiple geographic dimensions, and the load balancer selects agents based on spatial proximity to callers, reducing wide area network traffic while maintaining quality of service through local processing.
3Speed
If the system uses a large number of distributed media agents to reduce latency, then local response time improves, but managing available capacity and assigning agents becomes more complex
Solution Approach 1:
Media agents perform self-service by autonomously publishing their availability and location information to the load balancer. This eliminates the need for complex centralized management of agent assignments, as each agent independently announces its capabilities and the load balancer automatically routes requests based on this published information.
Solution Approach 2:
The system changes the management parameter from active assignment to passive advertisement. Instead of the system actively assigning agents to callers, media agents change their behavior to actively advertise their availability and location parameters, simplifying the management of distributed agents while maintaining fast local response times.
4Ease of operation
If the system maintains centralized control for conference session management, then coordination is simplified, but failure recovery time increases due to network separation delays
Solution Approach 1:
The load balancer performs preliminary actions by maintaining a pre-computed list of available media agents and their locations. When a failure occurs, the load balancer can immediately redirect traffic to alternative agents from its local cache without waiting for centralized coordination, enabling rapid failure recovery while maintaining simplified centralized control for normal operations.
Data Source
AI summary
A load balancer controls a cluster of media agents each configured to perform media packet processing operations in a conference session. The load balancer receives a Session Traversal Utilities for Network Address Translation/Translator (NAT) (STUN) request from a caller to establish media. The STUN request includes a conference identifier (ID) for the session. The load balancer determines an availability of each of the media agents in the cluster, selects a media agent for the caller based on the conference ID and the determined media agent availabilities, forwards the STUN request to the selected media agent, and routes media packets between the caller and the selected agent for the session.