Network Device Packet Processing via Classifier Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional network devices face challenges in implementing high-performance packet distribution capabilities, struggling to efficiently process and manage large quantities of end-to-end service flows due to limitations in their single-instance classifier architecture.
Innovation Solution
Deploying multiple software-instance classifiers and network service processing units within a network device, which allows for concurrent processing and distribution of packet identifiers and flow queue management, enhancing the packet distribution processing capability and supporting larger bandwidth scenarios.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a single-instance classifier is deployed in a conventional network device, then the device can process packets according to service flows, but it is difficult to implement high performance and high rate packet distribution capability
Solution Approach 1:
The patent divides the single-instance classifier into multiple software-instance classifiers (first classifier and second classifier). Each classifier independently processes packets from different queue areas, enabling parallel packet distribution operations. This segmentation transforms the single-threaded classifier into a multi-threaded system, directly improving packet distribution processing capability while maintaining manageable complexity through modular design.
2Productivity
If multiple software-instance classifiers are deployed for concurrent processing, then packet distribution processing capability is enhanced, but system complexity increases
Solution Approach 1:
The system segments the packet processing workload by creating dedicated queue areas (first queue area and second queue area) for each classifier. This spatial segmentation prevents classifiers from competing for the same resources, reducing coordination overhead. Each classifier operates independently on its assigned queue area, enabling concurrent processing without proportionally increasing system complexity.
Solution Approach 2:
The patent introduces a new dimension of parallelism by deploying multiple software-instance classifiers that operate simultaneously on different queue areas. This transforms the single-dimensional sequential processing into multi-dimensional concurrent processing, enhancing throughput while the modular architecture keeps complexity manageable through clear separation of concerns.
Data Source
AI summary
K classifiers and S network service processors are deployed in a network device. A network adapter receiving queue is divided into multiple queue areas. Each classifier may acquire a packet identifier from a queue area that corresponds to the classifier and is in the network adapter receiving queue, that is, the network adapter receiving queue has multiple information reading interfaces such that the K classifiers may concurrently read packet identifiers from corresponding queue areas in the network adapter receiving queue, and the K classifiers may further concurrently determine, based on the packet identifiers acquired by the K classifiers, corresponding flow queue identifiers. Related operations before the K classifiers distribute packet description information to flow queues may all be concurrently executed, and concurrent execution makes processing times of the related operations become partially or completely overlapped.