Loopback Packet Forwarding via Ingress Port Buffer Accounting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Information handling systems face inefficiencies in packet processing due to loopback configurations, which lead to reduced port bandwidth and internal head-of-line (HoL) blocking, causing performance hampering in network switches.
Innovation Solution
The proposed solution involves associating metadata with packets to track their original ingress port and buffer locations, allowing for efficient buffer accounting and loopback handling, thereby mitigating HoL blocking and optimizing resource usage by ensuring packets are accounted for correctly and reducing unnecessary congestion.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If loopback mode is enabled to allow packet recirculation through the NPU, then packet processing functionality is improved, but port bandwidth is reduced because fewer ports are available to receive external data traffic
Solution Approach 1:
The patent segments the port roles by designating specific ports as loopback ports (dedicated to recirculation) and non-loopback ports (dedicated to external traffic). This segmentation allows the system to maintain full port bandwidth for external traffic while enabling loopback functionality on dedicated ports, resolving the contradiction between packet processing adaptability and port bandwidth productivity.
2Adaptability or versatility
If loopback is used to extend packet processing stages, then processing capability is improved, but internal HoL blocking occurs which hampers system performance
Solution Approach 1:
The patent introduces a loopback handling component as an intermediary that sits between the ingress port and the packet processing pipeline. This intermediary identifies loopback packets, associates them with their original ingress port, and directs them through the processing pipeline without triggering HoL blocking signals to the original ingress port. This mediator resolves the contradiction by enabling extended processing capability while preventing the performance degradation caused by internal HoL blocking.
Solution Approach 2:
The patent inverts the traditional loopback approach by not simply recirculating packets through the same ingress port, but instead routing loopback packets through a dedicated loopback handling mechanism that recognizes their special status. This inversion prevents the HoL blocking problem that would normally occur in standard loopback configurations, allowing the system to maintain high performance while achieving extended packet processing capability.
3Adaptability or versatility
If loopback configuration is implemented to enable packet recirculation, then processing flexibility is improved, but buffer accounting becomes complex leading to congestion control issues
Solution Approach 1:
The loopback handling component acts as an intermediary that simplifies buffer accounting by maintaining awareness of the original ingress port for each loopback packet. This intermediary tracks buffer usage and congestion status from the perspective of the original ingress port, rather than treating loopback packets as new arrivals at the loopback port. This approach maintains processing flexibility while significantly reducing buffer accounting complexity compared to traditional loopback implementations.
Data Source
AI summary
Presented herein are systems and methods for packet processing in an information handling system that uses loopback processing. In one or more embodiments, buffer accounting is done with respect to the original ingress port for packets when using loopback processing. Such embodiments have several benefits, including but not limited to, mitigating or eliminating head-of-line (HoL) blocking of traffic and headroom (plus additional buffers which may need to be allocated for use before priority-based flow control is generated) provisioning for the loopback port. Embodiments may include additional features, such as handling anomalies (e.g., corruption) and reclaiming reserved buffers if corruption or delays occur.