Network Device Oversubscription Handling via Dynamic Packet Discard
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Network devices often experience congestion due to mismatched packet rates between the data path and control path, leading to increased latency and inefficiency, as the control path cannot handle the ingress packet rate of the data path, resulting in temporary or permanent congestion states.
Innovation Solution
Implementing a method within network devices to extract header information, determine classification identifiers, and selectively discard packets based on congestion states and discard thresholds, allowing for prioritization and burst absorption to manage packet flow and reduce latency by adjusting the data path's buffer and processing load.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the data path buffers all received packets until transmission is ready, then packet loss is reduced, but latency increases and the control path becomes overwhelmed
Solution Approach 1:
The patent changes the parameter of buffer management by introducing dynamic discard thresholds based on congestion state. Instead of buffering all packets indefinitely, the system adjusts the buffer retention parameter dynamically - when congestion is detected (control path cannot keep up with data path ingress rate), packets are discarded based on classification identifiers and configurable thresholds, thereby reducing latency while maintaining reliability for acceptable packets.
Solution Approach 2:
The patent applies partial action by selectively buffering only certain packets based on their classification identifiers and the current congestion state. Rather than buffering all packets (excessive action), the system buffers only those packets that meet the discard threshold criteria, partially processing the packet stream to balance reliability and latency requirements.
2Measurement precision
If the control path processes all packet headers, then routing accuracy is maintained, but the control path cannot handle the ingress packet rate, causing congestion
Solution Approach 1:
The patent applies preliminary action by having the data path perform packet classification and determine classification identifiers before packets reach the control path. This preliminary classification allows the control path to make informed decisions about which packets to process fully and which to discard, maintaining routing accuracy for selected packets while improving overall packet handling rate by filtering out unnecessary traffic early in the pipeline.
Solution Approach 2:
The patent applies local quality by treating different packets differently based on their classification identifiers. Instead of uniform processing of all packets (which would overwhelm the control path), the system applies different quality levels of processing - full header processing for high-priority packets that meet discard thresholds, and discarding for low-priority packets - thereby maintaining routing accuracy where needed while improving overall productivity.
3Loss of time
If packets are discarded based on congestion state, then latency is reduced and control path overload is prevented, but some packet loss occurs
Solution Approach 1:
The patent resolves this contradiction by making the discard decision parameter dynamic rather than fixed. The discard threshold is configured based on the congestion state and classification identifier, allowing the system to adjust the reliability-latency tradeoff in real-time. When congestion is severe, more aggressive discarding reduces latency; when congestion is mild, more packets are retained for reliable delivery.
Solution Approach 2:
The patent applies the discarding principle by selectively removing packets that would contribute to congestion and latency issues. By discarding low-priority packets during congestion states while maintaining processing for high-priority packets, the system recovers control path performance and reduces overall latency, accepting controlled packet loss as a tradeoff for improved system-wide performance.
Data Source
AI summary
Header information is extracted from a received packet by a data path portion of a network device. The data path portion is configured to buffer a data portion of received packets until the received packets are ready for transmission from the network device. The data path portion determines a first classification identifier for the received packet based on the header information. The data path portion determines a congestion state of the data path portion. The congestion state indicates a received packet rate of the first data path portion that exceeds a packet handling rate of a control path portion of the network device. The data path portion discards the first packet if the congestion state meets a discard threshold associated with the first classification identifier.