Network Switch Path Selection for Low Packet Latency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Ethernet network switches face high latency due to the store-and-forward approach, where packet switching begins only after the end-of-packet arrives, leading to inefficiencies in packet processing.
Innovation Solution
Implement a low-latency path in network switches that allows packet switching to commence upon the arrival of the start-of-packet, bypassing or parallelizing cell switching latency, and maintain performance through eligibility checks and separate paths for different traffic types.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If store-and-forward approach is used, then packet integrity can be ensured, but packet processing latency increases
Solution Approach 1:
The patent segments packet processing into two distinct paths: a low-latency path for eligible packets that enables cut-through switching, and a store-and-forward path for packets requiring integrity verification. This segmentation allows the system to achieve both low latency for priority traffic and reliability for critical packets without forcing the entire system to use the slower SAF approach.
Solution Approach 2:
The patent applies different processing qualities to different packet types at the same network infrastructure. Eligible packets (e.g., VoIP, video conferencing) receive cut-through processing with minimal latency, while other packets undergo full store-and-forward processing. This local differentiation of processing quality resolves the contradiction by matching processing depth to packet requirements.
2Loss of time
If cut-through switching is implemented, then packet processing latency is reduced, but packet loss may increase due to insufficient buffering
Solution Approach 1:
The patent segments packets into eligible and non-eligible categories, applying cut-through switching only to eligible packets that have guaranteed QoS parameters. Non-eligible packets continue through the store-and-forward path with full buffering. This segmentation prevents packet loss by ensuring cut-through is only used when buffer space and QoS guarantees are available.
Solution Approach 2:
The patent performs preliminary eligibility assessment of packets before committing them to the low-latency cut-through path. This preliminary action filters packets that would cause buffer overflow or QoS violations, preventing packet loss while maintaining low latency for suitable traffic types.
3Adaptability or versatility
If multiple paths are created for different traffic types, then quality of service is improved, but device complexity increases
Solution Approach 1:
The patent implements a universal eligibility assessment mechanism that handles multiple traffic types (VoIP, video conferencing, real-time data) through a single decision framework. The same eligibility criteria and path selection logic apply to all packet types, providing multi-functional QoS support without proportionally increasing complexity.
Solution Approach 2:
The patent performs preliminary classification and eligibility assessment at the packet ingress point, making path selection decisions before packets enter the switching fabric. This preliminary action simplifies the core switching architecture by pre-processing traffic characterization and QoS verification, reducing the complexity of real-time path selection.
Data Source
AI summary
A network switch includes an input port configured to receive a first data packet comprising a header of metadata and a payload of a plurality of cells. The network switch includes a controller configured to determine, using at least the metadata, an eligibility of directing the first data packet to a first path or otherwise directing the first data packet to a second path. The first path is characterized by a first latency. The second path is characterized by a second latency. The first latency is lower than the second latency. The network switch is configured to start storing the first data packet from the first path into a queue upon receiving a first cell of the payload, to manage the packet in the queue, to retrieve the cells from the queue to prepare a second data packet; and to transmit the second data packet upon identification of a destination port.