LDBC SNB Interactive workload implementations

Reference implementations for the Interactive workload of the LDBC Social Network Benchmark (specification). To get the reference implementations for the BI workload, visit the ldbc_snb_bi repository.

⚠️ The goal of the implementations in this to serve as reference implementations which other implementations can cross-validated against. Therefore, our primary objective was readability and not absolute performance when formulating the queries.

Directory layout

Some configuration files and scripts use relative paths to address the data generator's directory. Hence, it is recommended to clone the LDBC Data Generator and the LDBC implementations repositories next to each other and keep their original directory names. For example:

• ldbc
  • ldbc_snb_datagen_hadoop
  • ldbc_snb_driver
  • ldbc_snb_interactive

Implementations

Each project has its own README:

• Cypher (Neo4j) implementation
• SQL (PostgreSQL) implementation

The Sparksee implementation is maintained in a separate repository.

Queries

The queries in this repository are work-in-progress. If possible, please cross-validate your queries against multiple implementations.

Interactive

Complex reads

query	01	02	03	04	05	06	07	08	09	10	11	12	13	14
Neo4j (Cypher)	01	02	03	04	05	06	07	08	09	10	11	12	13	14
PostgreSQL	01	02	03	04	05	06	07	08	09	10	11	12	13	14

Short reads

query	01	02	03	04	05	06	07
Neo4j (Cypher)	01	02	03	04	05	06	07
PostgreSQL	01	02	03	04	05	06	07

Inserts

query	01	02	03	04	05	06	07	08
Neo4j (Cypher)	01	02	03	04	05	06	07	08
PostgreSQL	01/p 01/c 01/e 01/l 01/t 01/u	02	03	04/f 04/t	05	06/p 06/t	07/c 07/t	08

User's guide

1. Grab the driver source code from:

   git clone https://github.com/ldbc/ldbc_snb_driver

2. Install the driver artifact to the local Maven repository:

   cd ldbc_snb_driver
   mvn clean install -DskipTests

3. Navigate to the root of this repository and build it to generate the JAR files for the implementations:

   ./build.sh

4. For each implementation, it is possible to (1) create validation parameters, (2) validate against an existing validation parameters, and (3) run the benchmark. In the directory of the implemenetation, set the parameters according to your system configuration in the .properties file and run the driver with one of the following scripts:

   driver/create-validation-parameters.sh
   driver/validate.sh
   driver/benchmark.sh

⚠️ Note that if the default workload contains updates which are persisted in the database. Therefore, the database needs to be re-loaded between steps – otherwise repeated updates would insert duplicate entries.

For more details, on validating and benchmarking, visit the driver wiki.

Generating small test data tests

To generate small data sets, use scale factor 1 (SF1) with the persons and years set according to this template. Note that a certain number of persons (~250) is required for the parameter generation script to function correctly.

ldbc.snb.datagen.generator.scaleFactor:snb.interactive.1

ldbc.snb.datagen.generator.numPersons:250
ldbc.snb.datagen.generator.numYears:1
ldbc.snb.datagen.generator.numThreads:1

ldbc.snb.datagen.serializer.personSerializer:ldbc.snb.datagen.serializer.snb.interactive.<SerializerType>PersonSerializer
ldbc.snb.datagen.serializer.invariantSerializer:ldbc.snb.datagen.serializer.snb.interactive.<SerializerType>InvariantSerializer
ldbc.snb.datagen.serializer.personActivitySerializer:ldbc.snb.datagen.serializer.snb.interactive.<SerializerType>PersonActivitySerializer