Network Processor Core Failure Detection via Dedicated Health-Check
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing network apparatuses face challenges in reliably detecting failures in all processor cores, especially when user traffic is present, due to the difficulty in assigning health-check packets to specific cores and the non-programmable hardware logic of packet schedulers.
Innovation Solution
A network apparatus with a control unit that transmits health-check packets to each processor core, monitors their status based on packet processing statistics, and assigns these packets to all cores without relying on the packet scheduler, allowing for precise failure detection and recovery processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If health-check packets are transmitted using round robin mode through the packet scheduler, then packet distribution is simplified, but reliable failure detection in all processor cores cannot be achieved
Solution Approach 1:
The patent introduces a dedicated health-check packet transmission mechanism that acts as an intermediary between the control unit and processor cores, bypassing the hardware packet scheduler. This intermediary mechanism ensures that health-check packets are systematically assigned to each processor core regardless of the scheduler's round-robin behavior, thereby enabling reliable failure detection while maintaining operational simplicity.
Solution Approach 2:
The patent segments the packet transmission function into two independent paths: one for user data packets through the hardware packet scheduler using round-robin mode, and another for health-check packets through a dedicated transmission mechanism. This segmentation allows each path to optimize for its specific purpose without interfering with the other, ensuring both ease of operation and reliability.
2Productivity
If hardware logic is used for packet scheduling, then wire speed performance is achieved, but programmability for specific packet assignment is lost
Solution Approach 1:
The patent segments packet handling into two independent streams: user data packets processed by the hardware packet scheduler at wire speed using round-robin assignment, and health-check packets handled by a dedicated transmission mechanism that provides programmable assignment to specific processor cores. This segmentation preserves the high-speed performance of hardware logic while adding the flexibility needed for systematic health monitoring.
Solution Approach 2:
The control unit is designed with multi-functionality, serving both as the initiator of health-check packets and as the monitor of processor core responses. Additionally, the system universally handles both user data packets and health-check packets through the same network interface, while using different assignment mechanisms for each type, thereby achieving both speed and adaptability.
3Productivity
If health-check packets are mixed with user packets, then network resource utilization is improved, but failure detection accuracy deteriorates
Solution Approach 1:
The patent applies local quality by treating health-check packets and user data packets differently in terms of assignment methodology. While both packet types traverse the network interface, health-check packets receive specialized handling with deterministic assignment to specific processor cores, whereas user packets follow the general round-robin scheduling. This localized differentiation ensures failure detection accuracy is not compromised by the presence of user traffic.
Solution Approach 2:
The dedicated health-check packet transmission mechanism serves as an intermediary that isolates the health-monitoring function from the general packet scheduling process. This intermediary ensures that health-check packets are systematically distributed to all processor cores regardless of user packet patterns, maintaining measurement precision while coexisting with user traffic in the network.
Data Source
AI summary
To detect a failure in each processor core appropriately. It is provided a network apparatus for transferring a packet, comprising: a control unit; and a network processor including a plurality of processor cores, each configured to perform a process of transferring a packet input over a network. The control unit being configured to: transmit a packet to the network processor; acquire a packet processing status of each of the plurality of processor cores; and monitor a status of the each of the plurality of processor cores based on the acquired packet processing status.