This project evaluates 4 graph databases - Neo4j, AgensGraph, LightGraph and TigerGraph, based on the LDBC_SNB benchmark. Scripts and experimental results are included so you can reproduce the benchmark results following our guide.
You can use ldbc_snb_datagen to generate LDBC raw data. The params.ini is our configuration file, in which you can change the value of scaleFactor to generate datasets with different scales. In this evaluation, we generate 3 datasets with scale factor 1, 10 and 100.
The installation and evaluation method of each database is listed in the corresponding folder. Most scripts and query statements are obtained from their websites or staff.
Seeds folder contains our experimental parameters. Seeds_1, seeds_10, seeds_100 contains parameters used in micro queries(interactive update and short query) for different scale datasets. Seeds_macro contains parameters used in macro queries(interactive complex and business intelligence query), which are same for different datasets.
We conducted all experiments on a single machine with two 20-core processors Intel Xeon E5-2680 v2 2.80GHz, 96GB main memory, and 960G NVMe SSD, runnig Ubuntu 16.04.5 operating system. You can find our results in result-all.csv.
Ran Wang, Zhengyi Yang, Wenjie Zhang, and Xuemin Lin. "An Empirical Study on Recent Graph Database Systems". The 13th International Conference on Knowledge Science, Engineering and Management (KSEM 2020).
Please send to wangransei@gmail.com for any further questions.