Ring Structure Priority Queues Memory Cache Eviction

The patent segments the cache into multiple priority queues (first priority queue, second priority queue, etc.) based on access patterns. By dividing the cache evaluation into segmented priority levels, the system avoids traversing all data blocks uniformly, instead focusing evaluation on specific priority queues, thereby reducing overall processing bandwidth consumption while maintaining eviction correctness.

The patent applies local quality by treating different data blocks with different evaluation intensities based on their priority. High-priority blocks in the first priority queue receive focused evaluation and protection from eviction, while lower-priority blocks in subsequent queues undergo less intensive evaluation. This localized differential treatment reduces total processing bandwidth while ensuring important blocks are preserved.

The patent implements dynamic priority queue assignment where data blocks can transition between priority queues based on access patterns. The system dynamically adjusts which queue a block belongs to, allowing hot blocks to move to higher priority queues and cold blocks to lower priority queues. This dynamic reorganization optimizes cache performance by placing frequently accessed blocks in faster-access queues while minimizing unnecessary movements through threshold-based transitions.

The patent changes the priority parameter of data blocks based on access frequency and recency. By modifying the priority level parameter rather than physically moving all blocks frequently, the system optimizes cache performance through parameter adjustment. Blocks only undergo physical movement when priority thresholds are crossed, reducing redundant data movements and associated processing overhead.