Incremental Graph Query Updates for Large Datasets
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Searching large datasets with frequent updates is computationally intensive due to the need to re-run queries over the entire dataset, even for small changes, leading to high resource usage and time consumption.
Innovation Solution
A method to determine if a graph update is localizable or relatively bounded, allowing incremental updates to query results without re-running the query over the entire graph, by inspecting only affected nodes within a specified number of hops and updating data associated with these nodes, using techniques such as key word search, pattern matching via subgraph isomorphism, regular path queries, and strongly connected components.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the entire graph is searched repeatedly to obtain updated results, then the accuracy and recency of search results are improved, but the computational resource usage and time consumption increase significantly
Solution Approach 1:
The patent divides the graph into affected and unaffected portions based on the change propagation analysis. Instead of processing the entire graph, only the segmented affected portion (nodes within k hops from changed nodes) is processed, reducing computational resources while maintaining result accuracy.
Solution Approach 2:
The patent applies partial action by performing query processing only on the necessary subset of the graph (affected nodes within k hops) rather than the entire graph. This partial processing achieves the required result accuracy without the excessive computational cost of full graph processing.
2Loss of time
If the entire graph is searched repeatedly to obtain updated results, then the recency of search results is improved, but the time to complete the search increases significantly
Solution Approach 1:
The patent segments the graph processing task into affected and unaffected portions. By identifying and processing only the affected nodes (those within k hops from changed nodes), the execution time is dramatically reduced while still providing up-to-date results for queries that depend on the changed data.
Solution Approach 2:
The patent performs preliminary analysis to identify affected nodes and their scope (k hops) before executing the query. This preliminary action enables the system to prepare the minimal necessary processing scope in advance, reducing the actual query execution time while maintaining result recency.
3Reliability
If the search algorithm requires the entire dataset to be searched again regardless of update size, then the completeness of results is ensured, but the computational complexity increases with dataset size
Solution Approach 1:
The patent applies local quality by making the processing scope adaptive to the local impact of changes. The parameter k defines a local radius around changed nodes, creating different processing scopes based on the local propagation of effects. This ensures completeness for affected regions while avoiding unnecessary processing in unaffected regions, reducing overall computational complexity.
Solution Approach 2:
The patent introduces dynamics by making the processing scope flexible and adaptive rather than static. The k-hop radius can be adjusted based on the type of change, query characteristics, and graph properties. This dynamic approach ensures result completeness when needed while reducing computational complexity when changes are localized and have limited impact.
Data Source
Figure 1~2
Figure 3~4
Figure 5~6
AI summary
A mechanism of updating query results for a graph linking data in a computer system is disclosed. Results of the query on the graph linking data are received along with a change to the graph. The change to the graph is determined to be localizable or relatively bounded. Based on the determination of the localizable or relatively bounded change to the graph, the results of the query are updated based on the change to the graph without determining updated results of the query over the graph. This is accomplished by discovering nodes that are affected by the change to the graph, updating data associated with the affected nodes, and applying the updated data to the results of the query.