TigerGraph together with Amazon Web Services (AWS) and Advanced Micro Devices, Inc. (AMD) set a new standard for graph database performance at scale, completing the world’s largest LDBC SNB BI benchmark on a 36TB dataset. Here is the full disclosure report.
The largest companies in financial services, retail, healthcare and manufacturing have their most valuable data stored in graph databases. Top use cases include fraud detection (JP Morgan Chase), and customer journey and personalization (Intuit).
“Using graph-based (machine learning) features, our model shows an amazing improvement in detecting 50% more risk events (fraud) and also improving model precision by 50% (false positives) at the same time.“
Uri Lapidot, Senior Product Manager, Intuit
Here is the full disclosure report SF30k LDBC SNB-BI benchmark powered by AWS EC2 AMD EPYC instances.
For those unfamiliar with the LDBC SNB-BI benchmark, it is designed to test the performance of graph technologies in handling complex, aggregation and join-heavy queries that touch a large portion of the graph with micro batches of insert and delete operations. In other words, it puts graph technologies to the test to see how well they can meet enterprise knowledge graph requirements, including data size and scale, data ingestion rates, query response time, depth of analytics and total cost of ownership.
The LDBC SNB-BI benchmark is a bit different from traditional relational database OLAP benchmarks, like TPC-H and TPC-DS, in that it evaluates graph-shaped queries. These are the kinds of queries that analyze connection patterns, complex pattern matching, and the cheapest pathfinding between many sources and destinations. Essentially, it looks at how well graph technology can understand and make sense of connections within data.
TigerGraph showed industry leading performance in handling these types of queries.The 36TB size of the graph used in the benchmark is no small feat, spanning 73 billion entities and 540 billion relationships. TigerGraph’s ability to handle large and complex queries on such a massive scale is a testament to the scalability and efficiency of the underlying architecture.
But wait, it gets even better. Thanks to the software updates in TigerGraph combined with the latest generation AMD powered Amazon R6a EC2 instances, data loading and average bi-query performance were 5x and 2x faster than previous TG LDBC SNB BI unaudited benchmarks, respectively.
All in all, the results of the LDBC SNB-BI unaudited benchmark showcase TigerGraph’s impressive performance in handling complex, aggregation, and join-heavy queries on large graphs. If you’re in the market for a graph technology platform that needs to handle big data and provide valuable insights, consider exploring TigerGraph and check out some of our Starter Kits.