Dynamic Uplink Pinning for Virtualized Network Congestion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In virtualized computing and networking environments, traffic congestion on physical links can lead to bandwidth imbalances, causing some classes of service to exceed capacity while others are underutilized, potentially violating service level agreements and resulting in financial losses or customer penalties.
Innovation Solution
The formation of uplink groups from multiple physical links allows for dynamic bandwidth reallocation and VM interface re-pinning based on congestion detection and QoS hashing, ensuring even distribution of traffic across members to align with allocated bandwidth shares and service level agreements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If VM interfaces are pinned to specific physical links, then traffic distribution is simplified and predictable, but bandwidth imbalance and congestion occur when traffic patterns change
Solution Approach 1:
The patent implements dynamic VM interface pinning where VMs can be reassigned to different physical links based on real-time traffic conditions. The system continuously monitors bandwidth utilization and congestion levels, then dynamically repins VM interfaces to optimize traffic distribution. This dynamic approach allows the system to adapt to changing traffic patterns while maintaining QoS guarantees, resolving the contradiction between simple pinned assignment and reliable bandwidth allocation under varying conditions.
2Quantity of substance
If multiple physical links are combined into uplink groups, then bandwidth capacity increases, but congestion detection and bandwidth reallocation complexity increases
Solution Approach 1:
The patent segments the uplink group into multiple physical links and further divides traffic into different VM interfaces and QoS classes. Each segment can be independently monitored and managed for congestion conditions. The system tracks bandwidth utilization per link, per VM interface, and per QoS class, enabling granular congestion detection and targeted bandwidth reallocation. This segmentation approach reduces the complexity of managing overall congestion by breaking down the management task into smaller, manageable units.
Solution Approach 2:
The system implements continuous feedback mechanisms that monitor bandwidth utilization and congestion levels in real-time. When congestion is detected on a specific physical link or for a specific QoS class, the feedback loop triggers bandwidth reallocation actions. The system adjusts VM interface assignments and bandwidth allocations based on this feedback, creating a closed-loop control system that automatically responds to changing conditions without requiring manual intervention or overly complex decision-making logic.
3Ease of manufacture
If bandwidth is statically allocated to traffic classes, then QoS guarantees are simple to implement, but bandwidth imbalances cause some classes to exceed capacity while others are underutilized
Solution Approach 1:
The patent transitions from static to dynamic bandwidth allocation for traffic classes. The system continuously monitors actual bandwidth consumption by each QoS class and adjusts allocations in real-time. When a class exceeds its allocated bandwidth, the system identifies alternative classes with available capacity and redistributes bandwidth accordingly. This dynamic adjustment maintains QoS guarantees while optimizing overall bandwidth utilization efficiency, resolving the contradiction between simple static allocation and efficient dynamic utilization.
Solution Approach 2:
The system changes the bandwidth allocation parameter from fixed static values to dynamic values that adapt to actual traffic conditions. The allocation parameters are continuously adjusted based on monitored bandwidth consumption patterns, congestion levels, and QoS class requirements. This parameter change allows the system to maintain fair bandwidth distribution while preventing any single class from exceeding capacity, thereby improving overall productivity and bandwidth utilization efficiency.
4Reliability
If VM interfaces are dynamically reassigned based on congestion, then bandwidth distribution fairness improves, but traffic management complexity increases
Solution Approach 1:
The system uses feedback mechanisms to automatically detect congestion conditions and trigger bandwidth reallocation. The feedback loop monitors bandwidth utilization and congestion levels, then automatically adjusts VM interface assignments without requiring complex manual traffic management decisions. This automated feedback-driven approach ensures fair bandwidth distribution while reducing the operational complexity of traffic management by eliminating the need for manual intervention in congestion resolution.
Solution Approach 2:
The system implements self-service traffic management where the network automatically detects and resolves congestion conditions without external intervention. The bandwidth reallocation process is autonomous, with the system monitoring its own state and making adjustments based on predefined policies and algorithms. This self-service capability improves bandwidth distribution fairness while reducing traffic management complexity by eliminating the need for manual traffic engineering and intervention.
Data Source
AI summary
At a network element having a plurality of physical links configured to communicate traffic over a network to or from the network element, an uplink group is formed comprising the plurality of physical links, wherein the plurality of physical links comprise a first physical link and a second physical link. A plurality of classes of service are defined comprising a first class of service and a second class of service, wherein the first class of service and second class of service have bandwidth allocations on the first physical link. Traffic congestion is detected on the first physical link that exceeds a predetermined threshold for the first class of service. Traffic associated with one or more virtual machines associated with the first class of service on the first physical link is re-associated to the second physical link until the traffic congestion falls below the predetermined threshold.