Network Packet Scheduler for Aggregate Traffic Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional congestion control methods in IP internetworks focus on end-to-end communication, disregarding aggregate flow between and within networks, leading to inefficiencies in bandwidth utilization and increased resource requirements, especially in bandwidth-constrained systems with many interconnected nodes.
Innovation Solution
The implementation of a packet scheduler that adapts bandwidth utilization and transmission delay based on network conditions, using a packet-in-flight counter and rate limiting to optimize aggregate packet traffic flow, with a focus on round trip time and dynamic adjustment of packet-in-flight limits to manage congestion across multiple networks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional end-to-end congestion control methods are used, then communication between two nodes is maintained, but aggregate packet traffic flow between networks is not optimized leading to bandwidth underutilization
Solution Approach 1:
The patent introduces an intermediary congestion control mechanism that operates at the network level between multiple networks, rather than only at end-to-end node level. This intermediary layer monitors and controls aggregate packet traffic flow between networks, optimizing bandwidth utilization without requiring complex changes to existing end-to-end communication protocols.
Solution Approach 2:
The congestion control function is segmented into different levels: traditional end-to-end control between nodes and a new network-level aggregate flow control. This segmentation allows each layer to operate independently with appropriate complexity, where the network-level controller handles aggregate optimization without interfering with simpler end-to-end communication.
2Productivity
If packet-in-flight limits are increased to improve throughput, then more packets can be transmitted, but network congestion and burstiness increase
Solution Approach 1:
The patent implements a feedback mechanism where the network-level congestion controller monitors packet-in-flight counts and network conditions, dynamically adjusting the maximum packet-in-flight limits. When congestion is detected, limits are reduced; when network capacity is available, limits are increased. This feedback loop maintains throughput optimization while preventing excessive congestion and burstiness.
Solution Approach 2:
The maximum packet-in-flight limits are made dynamic rather than static, allowing the system to adapt to changing network conditions. The controller adjusts these limits in real-time based on monitored traffic patterns and network state, enabling throughput optimization without causing sustained congestion or burstiness.
Data Source
AI summary
Systems and methods of optimizing packet flow in a network are disclosed. An internetwork has a first network and plural second networks. A respective packet-in-flight counter is established for each one of the plural second networks and a respective maximum packet-in-flight indication for each one of the plural second networks. The respective packet-in-flight counter for each one of the plural second networks when a request packet is sent from any node in the first network to any node in that one of the plural second networks, or a response packet is received from any node that one of the plural second networks. it is determined whether a second request packet can be sent from any node in the first network to any of the plurality of nodes in one of the plural second networks, based on processing a value of the respective packet-in-flight counter for that one of the plural second networks and a value of the respective maximum packet-in-flight indication for that one of the plural second networks. The respective maximum packet-in-flight indication for each one of the plural second networks is adjusted based on a statistic that is indicative of roundtrip time for request packets sent from any node of the first network to any of the plurality of nodes in that one of the plural second networks.