Graph database market leader Neo4j Inc. today announced new features that it says can improve the speed of analytical queries up to 100-fold and enable users to run both transactional and analytical workloads within a single database while tracking real-time data changes.

The performance is achieved via parallelization, which adds concurrent threads across multiple CPU cores to run analytical graph queries. Neo4j also leverages a technique called morsel-based parallelism that divides a workload into smaller pieces called morsels and processes them in parallel for better scalability and resource utilization.

Performance improvements “are a function of how big the query is and how much compute you bring to it,” said Sudhir Hasbe, Neo4j’s chief product officer. “If you have a four-core machine, it will be a four times improvement.”

The database also now includes change data capture, a technique that captures changes made to data — such as insertions, updates and deletions — so they can be transferred to other systems. CDC is widely used in replication scenarios in which updates are fast and frequent. The capability is also integrated with the Neo4j Connector for Apache and Confluent Inc.’s Kafka.

“Many of our customers use us as a system of record” for real-time fraud detection and understanding the relationship between parts in complex manufacturing scenarios, Hasbe said. “As soon as something changes with the graph dataset, they want to get an event and post-process it.”

Neo4J already had real-time updates and ingestion, “but this is the first time we are sending events in real-time so you can track the changes and build a different application,” Hasbe said. “One part of the business may have the graph, and another can monitor for and model changes.”

Another feature provides for easier Knowledge Graph creation via embedding models that predict and find missing relationships and infer new connections within an organization’s knowledge graph for greater semantic understanding.

Machine learning algorithms can find incomplete relationships. For example, “If you have been buying clothes for something from a digital retail store consistently, it can predict the kids’ age or how many kids you have,” Hasbe said. “It can automatically figure out relationships you don’t have to create up front.”

Two new pathfinding algorithms have also been added to identify critical paths and convert an unsorted graph into a topologically sorted graph with dependencies understood. The features build upon 65 algorithms already within the database, Hasbe said.

The new features will be included at no additional cost for customers with self-managed installations.

Image: Neo4j