Dynamic Load Balancing via Congestion Marking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing network load balancing methods, such as those defined in the IEEE 802.3ad standard, do not effectively distribute traffic load across multiple Ethernet links, leading to underutilization of bandwidth and potential packet out-of-order reception due to variable delays and unequal link capacities.
Innovation Solution
A dynamically load-balanced network switch architecture that uses congestion management mechanisms like random early detection (RED) and backward congestion notification (BCN) to distribute packets across physical links, ensuring packets are delivered in order and adaptively rebalance traffic based on changing load conditions, using techniques like weighted random early detection and last-packet marking to manage congestion and prevent packet duplication.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If packets are distributed across multiple elementary links using predetermined hashing procedure, then bandwidth utilization is improved, but packet delivery order is compromised due to variable delays on different links
Solution Approach 1:
The system performs preliminary actions by marking the last packet of each conversation before distribution and buffering packets at the ingress transmission point. This preliminary marking and buffering ensures that when packets are distributed across multiple links, the receiver can reconstruct the original order by identifying the marked last packet and holding buffered packets until proper sequence is achieved.
Solution Approach 2:
The patent introduces an intermediary mechanism through the use of marked packets and buffering at the ingress point. The mark acts as an intermediary signal that travels with the last packet, and the buffer serves as an intermediary storage that temporarily holds packets to maintain order. This intermediary approach allows the system to distribute packets across multiple links while preserving delivery order at the receiver.
2Ease of operation
If traffic load is allocated based on predetermined hashing, then distribution simplicity is maintained, but link load balancing efficiency deteriorates due to unequal link capacities and priorities
Solution Approach 1:
The patent implements dynamics by transitioning from static predetermined hashing to dynamic load balancing. The system continuously monitors link status, packet queues, and congestion levels, then dynamically adjusts packet distribution decisions. This dynamic approach allows the system to adapt to changing link capacities and traffic priorities, improving load balancing efficiency while maintaining operational simplicity through automated control.
Solution Approach 2:
The system employs feedback mechanisms by monitoring link congestion status, queue lengths, and traffic patterns, then using this information to adjust packet distribution. The feedback loop continuously evaluates link performance and modifies routing decisions accordingly, enabling the system to optimize load balancing efficiency across links with unequal capacities and priorities while maintaining simple operation through automated feedback-driven control.
3Productivity
If link aggregation is implemented to increase throughput, then bandwidth capacity is improved, but system complexity increases due to coordination and congestion management across multiple links
Solution Approach 1:
The patent applies self-service by enabling each transmission point to autonomously perform congestion detection, packet marking, and load balancing decisions without requiring complex centralized coordination. The system uses distributed intelligence where each node independently monitors its own queues and links, making local optimization decisions. This self-service approach increases bandwidth capacity while reducing coordination complexity by eliminating the need for complex inter-node communication and centralized control.
Data Source
AI summary
Apparatus, systems, methods, and articles described generally herein may receive a first packet marked with a congestion indicator (CI). Upon receipt of the CI, a load-balancing operation may be performed among a plurality of physical links upstream from a point of congestion to alleviate the congestion. Other embodiments may be described and claimed.