Graph Database Benchmark Reports

Graph Database Benchmark Reports


Linked Data Benchmark Council Social Network Benchmark (36TB raw dataset)

In a recent study, we measured TigerGraph’s performance using the respected Linked Data Benchmark Council (LDBC) Social Network Benchmark (SNB) Scale-Factor 30K dataset  (73B vertices, 534B edges, 36TB raw data).

We focused primarily on Business Intelligence (BI) workload tests during this benchmark: BI Workload — The majority of OLAP-style iterative and deep-link graph queries were answered in under a few minutes.

The data schema follows the property graph data model. Measurements during the benchmark, which we believe to be the first benchmark test using the LDBC-SNB SF-30K BI workload on a distributed graph database, included loading time, storage size, and query latency of the 20 BI queries on a cluster of 40 machines.

This benchmark test demonstrates TigerGraph’s capability handling large-scale updateable connected data with a set of demanding graph benchmark queries. To date, no other graph database demonstrated a comparable result.

GitHub Location with datasets as well as procedure to reproduce

Benchmark Feb 2022 Cover

Third Party Benchmark:
Neo4j vs TigerGraph

July 2019 In-Depth Benchmarking of Graph Database Systems using the Linked Data Benchmark Council (LDBC) Social Network Benchmark (SNB).  

Graph Database Benchmark Report

This benchmark compares TigerGraph to Neo4j, Amazon Neptune, JanusGraph and ArangoDB in the following categories:

Queries: For 2-hop path queries, TigerGraph is 40x to 337x faster than other graph databases.

Storage: When comparing storage requirements for the same raw data, other graph databases need5x to 13x more disk space compared to TigerGraph.

Loading: TigerGraph loads data1.8x to 58x faster than other graph databases.

Scale: TigerGraph scales almost linearly with additional machines, achieving 6.7x speedup with 8 machines for the computationally intensive PageRank algorithm. Neo4j and Amazon Neptune cannot partition a graph across multiple machines for the versions tested, thus could not be tested for this part.

Click below for the benchmark summary or download the full report.