Multi-Lane Link Error Detection Using Per-Lane CRC Replay
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current Ethernet protocols and InfiniBand link-layer implementations lack inherent facilities for reliable data transmission, leading to increased processing overhead and latency due to the need for explicit ACKs, which consume link bandwidth and add complexity, especially in high-performance computing environments where data transfer latencies are a significant bottleneck.
Innovation Solution
The proposed architecture employs a lane error detection and removal mechanism using implicit ACKs with a replay buffer and per-lane CRCs to ensure reliable data transmission across multi-lane links, reducing the probability of data corruption and improving link reliability without explicit acknowledgments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If explicit ACKs are used for reliable data transmission, then data reliability is improved, but link bandwidth is consumed and processing overhead increases
Solution Approach 1:
The patent extracts the reliability guarantee function from the link layer and moves it to the network layer. The link layer simply forwards packets without performing ACK operations, while the network layer (TCP) handles all reliability mechanisms including ACKs and retransmissions. This separation eliminates link bandwidth consumption by ACKs while maintaining end-to-end reliability.
Solution Approach 2:
The patent introduces a bridge device as an intermediary that can perform ACK generation and packet retransmission functions. The bridge acts as a mediator between the link layer and network layer, providing reliable delivery without requiring end systems to consume additional link bandwidth for ACKs, as the bridge handles acknowledgments locally.
2Reliability
If explicit ACKs are used for reliable data transmission, then data reliability is improved, but latency increases
Solution Approach 1:
By extracting the reliability function from the link layer to the network layer, the patent eliminates the time required for link-layer ACK exchanges. Packets are forwarded immediately by the link layer without waiting for acknowledgments, reducing latency while TCP at the network layer maintains reliability through its own ACK and retransmission mechanisms.
Solution Approach 2:
The patent implements preliminary buffering at the bridge device, where packets are pre-stored before being forwarded. This allows the bridge to quickly retransmit packets if needed without introducing additional latency, as the packets are already in memory ready for immediate re-sending, eliminating the need for time-consuming retransmission requests and waits.
3Reliability
If per-packet ACKs are used for reliable data transmission, then data integrity is improved, but processing complexity increases
Solution Approach 1:
The patent extracts the complexity of reliability management from the link layer network stack and concentrates it at the network layer (TCP). The link layer becomes a simple forwarding entity without ACK generation, sequence tracking, or retransmission logic, significantly reducing its processing complexity while TCP handles all integrity guarantees at the end-to-end level.
Solution Approach 2:
The bridge device acts as an intermediary that simplifies the network stack by taking over ACK generation and retransmission functions. This allows end systems to use simpler link layer protocols while the bridge provides the necessary reliability mechanisms, reducing processing complexity at the end systems without sacrificing data integrity.
4Reliability
If buffer space is maintained for packet retransmission, then data reliability is improved, but memory resources are consumed
Solution Approach 1:
The bridge device serves as an intermediary that provides retransmission capability without requiring end systems to maintain large buffers. The bridge stores packets locally in its forwarding buffer, enabling retransmission if packets are lost or need to be resent, while end systems can use smaller buffers since the bridge absorbs the retransmission burden.
Data Source
Figure 1
Figure 2~13
Figure 3~4
AI summary
Method, apparatus, and systems for detecting lane errors and removing errant lanes in multi-lane links. Data comprising link packets is split into a plurality of bitstreams and transmitted over respective lanes of a multi-lane link in parallel. The bitstream data is received at multiple receive lanes of a receiver port and processed to reassemble link packets and to calculate a CRC over the data received on each lane. The link packets include a transmitted CRC that is compared to a received CRC to detect link packet errors. Upon detection of a link packet error, per-lane or per transfer group CRC values are stored, and a retry request is issued to retransmit the bad packet. In conjunction with receipt of the retransmitted packet, per-lane or per transfer group CRC values are recalculated over the received data and compared with the stored per-lane or per transfer group CRC values to detect the lane causing the link packet error.