Azure HorizonDB introduces substantial Apache AGE performance improvements for querying and building knowledge graphs. Enhanced query planning and optimized graph joins accelerate eligible read queries, while faster duplicate detection and improved memory management make ingestion using `MERGE` more efficient.
Benchmark results include:
- Up to 3,900× faster short
graph lookups: In LDBC SNB tests with approximately 9 million vertices and 53
million edges, listing a person’s friends improved from approximately 1.8
seconds to 0.46 milliseconds. Selected complex queries improved by up to 512×.
- About 190× faster `MERGE`
ingestion: Processing one million inputs with 100,000 distinct vertices
improved from 906 seconds to 4.8 seconds.
- 83% lower sampled peak
backend memory usage: In a separate 10-million-input test, memory usage fell
from approximately 65.2 GiB to 11.0 GiB.
Both read-query optimizations are enabled by default.
Statistics-aware property estimates require explicitly creating property
statistics and running `ANALYZE`. Performance gains vary by query, dataset, and
configuration.
Learn more.