Buffer Memory Management for Read Speed Optimization

The patent segments the mapping table search process by introducing a buffer management mechanism that divides data into buffer units with associated mapping entries. When the buffer contains many data units, the system segments the search space by checking buffer hits first (O(1) access) before falling back to full mapping table search, effectively breaking down the large-scale search problem into smaller, manageable segments.

The patent performs preliminary actions by pre-establishing the logical unit-buffer unit mapping table before read operations occur. This mapping table is built in advance during data loading, so when read requests arrive, the system can quickly check the pre-computed mapping relationships without performing complex searches during the actual read operation, reducing latency.

The patent implements dynamic buffer management where the buffer memory capacity and mapping table size adjust based on workload conditions. The system dynamically allocates buffer units, updates mapping entries selectively, and can evict less frequently accessed data, allowing the mapping table size to remain manageable even as overall buffer capacity increases to serve more requests.

The patent changes key parameters including buffer unit size, mapping table organization structure, and data eviction thresholds based on system conditions. By adjusting these parameters, the system optimizes the balance between buffer capacity (for serving more requests) and mapping table complexity (for maintaining fast access), adapting to different workload scenarios.