Logarithmic Access Counters for Automated Database Index Advising

The patent transforms the counting mechanism from linear to logarithmic scale. Instead of counting exact access frequencies, the system uses logarithmic counters that record the order of magnitude of access frequencies. This parameter transformation maintains sufficient precision for database tuning decisions while dramatically reducing the memory footprint, as logarithmic values require fewer bits to represent the same range of frequencies.

The patent implements different counting strategies for different data structures and access patterns. Logarithmic counters are used specifically for index advisor access counters where precise frequency ordering is needed but exact counts are unnecessary. This localized application of logarithmic counting optimizes memory usage for the specific purpose of index recommendation without compromising the quality of workload statistics for that particular function.

The system changes the parameter being measured from exact access counts to logarithmic frequency buckets. During query execution, instead of incrementing precise counters for each access, the system determines which logarithmic bucket the access frequency falls into and increments that counter. This approach provides sufficient precision for determining relative access frequencies while minimizing the computational overhead during query execution.

The patent employs lazy counter creation where logarithmic counters are only created when actually needed during query execution, rather than pre-allocating all possible counters. This on-demand creation reduces the initialization overhead and memory allocation time, making the statistics collection process more efficient while maintaining measurement precision when counters are actively used.

The patent creates a universal logarithmic counter framework that serves multiple advisor tools (index advisor, data compression advisor, buffer pool size advisor, table partitioning advisor). The same logarithmic counter infrastructure and measurement mechanisms are reused across all advisor types, reducing overall system complexity despite the comprehensive coverage of multiple advisors. Each advisor leverages the same core counting technology adapted to its specific needs.

The system segments the counter management into distinct components: logarithmic counter data structures, lazy creation mechanisms, and advisor-specific access patterns. This segmentation allows each component to be optimized independently while maintaining overall system coherence. The modular structure reduces complexity by clearly defining interfaces and responsibilities between different parts of the counter management system.