Fragment-Based Interrupt Coalescing for Network Controllers
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional interrupt coalescing methods in network controllers for Ethernet packets fail to preserve latency for high-priority packets, particularly in safety-critical applications like autonomous driving, as they coalesce interrupts based on fixed parameters without considering packet fragmentation, leading to delays in processing critical data.
Innovation Solution
The network controller employs a fragment-based interrupt coalescing engine that determines whether to coalesce interrupts based on the last fragment index and total number of fragments in Ethernet packets, specifically for protocols like RTPS and SOME/IP, ensuring that interrupts for high-priority packets are not delayed, thereby reducing processor load while maintaining low latency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional interrupt coalescing methods are used with fixed parameters, then processor load is reduced, but latency for high-priority packets increases
Solution Approach 1:
The patent segments the interrupt coalescing logic into two distinct paths: one for fragmented packets and one for non-fragmented packets. This segmentation allows the system to apply different coalescing strategies based on packet type, thereby reducing processor load for fragmented traffic while maintaining low latency for high-priority non-fragmented packets through the separate fast path.
Solution Approach 2:
The patent implements dynamic interrupt coalescing by adjusting coalescing behavior based on real-time packet characteristics. The system dynamically switches between coalescing and non-coalescing modes depending on whether packets are fragmented or not, and based on priority levels, allowing optimal balance between processor load reduction and latency preservation under varying traffic conditions.
2Device complexity
If interrupt coalescing is performed without considering packet fragmentation, then interrupt processing is simplified, but safety-critical applications experience delays
Solution Approach 1:
The patent segments interrupt handling into distinct pathways based on packet fragmentation status. One pathway handles fragmented packets with appropriate coalescing logic, while another pathway handles non-fragmented packets with guaranteed low-latency processing. This segmentation maintains reliability for safety-critical applications by ensuring they traverse the fast path without unnecessary coalescing delays.
Solution Approach 2:
The patent introduces an intermediary classification mechanism that examines packet headers to determine fragmentation status and priority level. This intermediary layer directs packets to appropriate processing queues, ensuring that safety-critical non-fragmented packets are immediately forwarded to the host processor while allowing other traffic to undergo coalescing, thus maintaining both simplicity and reliability.
3Device complexity
If fixed-parameter coalescing is applied to all packets, then implementation is simpler, but high-bandwidth sensors cause excessive processor interrupts
Solution Approach 1:
The patent segments the interrupt coalescing implementation into protocol-specific handlers, particularly for high-bandwidth sensor protocols like RTPS and SOME/IP. This segmentation allows the system to apply aggressive coalescing specifically to fragmented sensor traffic while maintaining standard handling for other protocols, thereby reducing overall interrupt frequency without requiring complete redesign of the coalescing architecture.
Solution Approach 2:
The patent changes coalescing parameters dynamically based on packet type and protocol identification. For high-bandwidth sensor protocols with fragmented packets, the system adjusts coalescing thresholds and timing parameters to maximize interrupt reduction. This parameter adaptation allows simple implementation for common cases while providing optimized performance for specific high-traffic protocols.
Data Source
AI summary
A packet is transmitted from a remote device over a communication network. A fragment detector detects one or more fragments in a field of the packet, where the field is associated with a session layer or higher abstraction layer of an open systems interconnect (OSI) model. Fragment information is extracted from the packet which indicates one or more of a last fragment index associated with a last fragment of one or more fragment in the packet and a fragment count indicative of a number of fragments associated with a message which is fragmented. Interrupts associated with the packet with other interrupts associated with other packets are coalesced based on one or more of the last fragment index and the fragment count.