Database Index Contention Under High Concurrency

The patent segments the index update operations by dividing the index structure into multiple segments or ranges. Instead of updating the entire index sequentially with monotonically increasing values, the system divides index updates into parallel segments that can be processed concurrently, thereby maintaining data retrieval accuracy while improving write throughput during high concurrency periods.

The patent introduces a new dimension to index value management by using non-monotonic index values that incorporate multiple dimensions (such as combining sequential counters with hash-based distributions or using multi-level indexing). This allows the system to maintain accurate data retrieval by preserving logical ordering while enabling parallel updates through dimensional decomposition of the index space.

The patent implements a universal index update mechanism that handles both high concurrency write operations and accurate data retrieval through a single integrated approach. The system uses a multi-functional index structure that simultaneously supports parallel updates and efficient queries without requiring separate sharding infrastructure, thereby improving write throughput while avoiding the complexity of multiple specialized systems.

The patent uses copying techniques by maintaining multiple copies or representations of index data in different formats or locations. Instead of physically sharding the database, the system creates virtual copies of index structures that can be updated in parallel while presenting a unified interface to users, thus improving throughput without increasing actual system complexity.

The patent replaces the mechanical sequential ordering mechanism with a computational approach. Instead of relying on physical monotonic ordering of index values for query performance, the system uses computational algorithms (such as hash functions, bitmap indexes, or inverted indexes) that can handle non-monotonic values while maintaining efficient query performance through software-based optimization rather than mechanical ordering.

The patent changes the parameters of the index structure by allowing index values to be non-monotonic while compensating through additional parameters or metadata. The system modifies the index to include auxiliary information (such as position markers, range boundaries, or transformation functions) that enables efficient queries even when the primary index values are distributed non-monotonically, thus improving write throughput without sacrificing query performance.