Packet Switch Congestion Management via Bandwidth-Based State Purging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing packet switching networks face inefficiencies in managing congestion within switch elements due to limitations in buffer resources, leading to potential data loss and poor performance, particularly in managing the release of halted packets when congestion conditions cease.
Innovation Solution
A system and method that buffer packets in virtual output queues, halt the sending entity upon congestion, store the halt condition in switch states, identify the path occupying the greatest portion of the queue, update bandwidth indicators, and purge the state with the smallest bandwidth to efficiently manage congestion and resource usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If flow turn-on is accomplished through a timer at the source, then the halt state can be reset after a time interval, but this results in inefficient usage of switching resources and poor overall performance of the node
Solution Approach 1:
The patent implements feedback by having the switch element monitor queue congestion status and send release signals (XON) back to sources when congestion ceases. This allows dynamic adjustment based on actual network conditions rather than fixed timer intervals, improving both reliability and productivity by releasing packets only when the network is ready to handle them.
Solution Approach 2:
The switch element autonomously monitors its own queue congestion status and manages the release of halted packets without external control. The switch element maintains halt states internally and automatically sends release signals when appropriate, enabling self-service congestion management that improves resource utilization and throughput.
2Productivity
If the switch element monitors congestion and sends release signals to sources, then packet release can be optimized, but there are a large number of paths and it is difficult to manage all the paths due to physical and/or cost constraints
Solution Approach 1:
The patent merges the management of multiple paths by having the switch element aggregate congestion monitoring and release signal generation for all paths passing through it. Instead of managing each path independently at multiple points, the switch element consolidates control, reducing overall system complexity while maintaining throughput optimization.
Solution Approach 2:
The switch element acts as an intermediary between multiple sources and destinations, centralizing the congestion management function. By positioning the monitoring and control logic at the switch element rather than at each source or destination, the system reduces the number of monitoring points needed while maintaining effective path management.
3Productivity
If buffers are arranged in ingress parts at the inputs of switch elements, then data throughput can be maintained, but congestion and loss of data may occur due to limitations in the number of viable buffers
Solution Approach 1:
The patent applies preliminary action by halting packet transmission at the source before the buffers become completely full. Flow control mechanisms stop packets at the source when queue thresholds are reached, preventing buffer overflow and data loss before it occurs. This proactive approach maintains reliability while allowing the buffers to be used efficiently for throughput.
Solution Approach 2:
The system uses feedback through flow control mechanisms that monitor buffer occupancy and send halt signals (XOFF) to sources when congestion is detected. This feedback loop prevents data loss by stopping transmission before buffers overflow, while still allowing maximum throughput when buffers have available capacity.
Data Source
AI summary
A system and method are disclosed for switching a data flow of information packets intended for paths between a respective sending and receiving entity, the method includes buffering the packets from the paths in a queue; halting a sending entity on congestion of the queue; storing the halt condition in a switch state; noting the individual portions that different of the paths occupy in the queue; halting the sending entity for the path occupying the individually greatest portion of the queue; storing the halted path in a free one of the switch states including storing its bandwidth; successively updating the respective bandwidth of halted paths as the queue is repeatedly congested; determining an older part of the states; and purging the state for a path having the smallest bandwidth in said older part of the states.