Dynamic Packet Classification via Parallel Sub-Classifiers

The patent divides a single packet classification task into multiple sub-classification tasks that can be processed in parallel by multiple classifiers. Each classifier handles a subset of packets or a specific portion of the classification problem, allowing the system to achieve higher aggregate processing throughput without requiring each individual classifier to have excessive computing power. This segmentation enables cost-effective scaling by distributing the workload across multiple moderate-power classifiers rather than concentrating it in one high-power classifier.

The patent implements dynamic configuration where classifiers can be instantiated, activated, or deactivated based on real-time workload conditions. In cloud and virtualized environments, the system can scale the number of classifiers up or down according to actual packet processing demands, rather than maintaining a fixed over-provisioned configuration. This dynamic approach ensures reliability during peak demand while improving hardware utilization during lower-demand periods by activating only the necessary number of classifiers.

The patent segments the packet classification workload into multiple parallel processing streams, each handled by a separate classifier. Instead of one complex sequential classifier, the system uses multiple simpler classifiers working simultaneously on different packet subsets. This segmentation maintains relative simplicity in each individual classifier while achieving high aggregate throughput through parallelism, effectively resolving the trade-off between device complexity and productivity.