Hash Engine Index Separation for Point Query Latency

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Point queries in database systems and cloud-based storage often result in high tail latency due to read input/output (I/O) operations and the need to access disks, particularly in systems using solid-state drives (SSDs).

Innovation Solution

The implementation of a database system that generates a compact index from index information, which is separate from the data component, using compression and omission techniques. This compact index is stored in a cache memory, allowing for efficient point queries by reducing the number of disk I/O operations required.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Speed

If index information is stored in cache memory to speed up point queries, then query speed is improved, but cache memory size requirements increase

Engineering Contradiction:
Improvepoint query speedVSAvoidcache memory size
Core Design Contradiction:
SpeedVSVolume of stationary object

Solution Approach 1:

The patent extracts only the essential index information from the complete index structure and stores it in cache memory. Specifically, it separates the index into different components, caching only the most frequently accessed portions (such as the upper levels of the B-tree or the most recently used index entries) while leaving less frequently accessed index data on disk. This selective extraction allows the system to achieve fast query performance for common operations without requiring the entire index to reside in cache memory.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent implements a nested index structure where multiple levels of index information are organized hierarchically. The cache memory stores nested levels of the index structure, with higher-level index entries (which provide broader navigation) stored in cache, while lower-level detailed index entries remain on disk. This nested organization allows the system to maintain a compact cached index that still provides efficient query routing, as the cached higher-level indices can quickly identify which disk-resident lower-level indices need to be accessed.

Inventive Principle:
Principle #7Nested doll (Nesting)

2Measurement precision

If complete index information is cached, then query accuracy is improved, but memory usage increases

Engineering Contradiction:
Improvequery accuracyVSAvoidmemory usage
Core Design Contradiction:
Measurement precisionVSQuantity of substance

Solution Approach 1:

The patent applies local quality by ensuring that the cached index information is precisely tailored to meet the specific query requirements rather than caching all possible index data. The system analyzes query patterns and caches index entries that are locally optimal for those patterns, maintaining high query accuracy for the actual workload while using memory efficiently. This means different parts of the index structure are cached based on their specific utility for different types of queries, rather than uniformly caching everything.

Inventive Principle:
Principle #3Local quality

3Quantity of substance

If index information is compressed to reduce cache size, then memory efficiency is improved, but index retrieval complexity increases

Engineering Contradiction:
Improvecache memory capacityVSAvoidindex processing complexity
Core Design Contradiction:
Quantity of substanceVSDevice complexity

Solution Approach 1:

The patent changes the parameters of the index structure to achieve better compression characteristics. Specifically, it transforms the traditional B-tree index into a format that is more amenable to compression, such as by encoding key values more efficiently, removing redundant information, or reorganizing the data layout. These parameter changes allow the index to be stored in a compressed state in cache memory while still enabling efficient retrieval through appropriately designed decompression or interpretation routines.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP4530878A1Hash engine for conducting point queries
Publication Date: 2025.04.02 LEMON INC(GB)
  • EP4530878A1 patent drawingFigure 1
  • EP4530878A1 patent drawingFigure 2
  • EP4530878A1 patent drawingFigure 3

AI summary

Systems and methods are provided for improved point querying of a database. The index values are separated from data and retained in cache memory to allow access without requiring a disk input/output (I/O) operation and thereby having less latency resulting from such disk I/O operations.