LDBC SNB Interactive workload

The LDBC SNB Interactive workload captures transactional graph processing scenario with complex read queries that access the neighbourhood of a given node in the graph and update operations that continuously insert new data in the graph. Its data sets are available in the CWI/SURF data repository.

The workload produces the throughput metric to characterize the performance at the given scale. The full disclosure reports (FDRs) further detail the performance of the system under test by listing the data loading time and single query execution times.

For an overview of the workload, see the SIGMOD 2015 paper and the 2024 presentation by Gábor Szárnyas. For a detailed specification, see the LDBC Social Network Benchmark specification.

The tables below contain audited results. Note that the system cost indicates the 3-year TCO (total cost of ownership), which is the sum of the license, hardware, and maintenance costs (where maintenance means 24/7 support with a response time of less than 4 hours).

Audited results using a declarative query language

Benchmark setup	SF	Hardware	Performance	Performance (price-adj.)	Documents
System: Graph Engine Service 3.0.0 Test sponsor: Huawei Cloud Date: 2024-09-16 Query language: Cypher System cost: 3,105,819 RMB (437,840.04 USD)	30	Huawei Cloud ac8.48xlarge.4 96×AMD 2.4GHz vCPUs, 768GiB RAM	37,775.91 ops/s	86.278	Full disclosure report Executive summary Signatures Supplementary package
	100	Huawei Cloud ac8.48xlarge.4 96×AMD 2.4GHz vCPUs, 768GiB RAM	45,432.22 ops/s	103.764
	300	Huawei Cloud ac8.48xlarge.4 96×AMD 2.4GHz vCPUs, 768GiB RAM	40,347.04 ops/s	92.15
System: GraphDB 10.1.1 Test sponsor: Ontotext Date: 2023-01-31 Query language: SPARQL System cost: 216,222 USD	30	AWS r6id.8xlarge 32×Intel Xeon Platinum 8375C vCPUs, 256GiB RAM, 1R/1W thread	3.04 ops/s	0.014	Full disclosure report Executive summary Signatures Supplementary package
	30	AWS r6id.8xlarge 32×Intel Xeon Platinum 8375C vCPUs, 256GiB RAM, 2R/2W threads	6.76 ops/s	0.031
	30	AWS r6id.8xlarge 32×Intel Xeon Platinum 8375C vCPUs, 256GiB RAM, 4R/4W threads	12.16 ops/s	0.056

Audited results using an imperative language

Benchmark setup	SF	Hardware	Performance	Performance (price-adj.)	Documents
System: GraphScope Flex 0.26.1 Test sponsor: Alibaba Cloud Date: 2024-05-14 Queries implemented in: C++ System cost: 738,724 RMB (102,128.22 USD)	100	Alibaba Cloud ecs.r8a.16xlarge 64×AMD EPYC 9T24 @ 3.7GHz vCPUs, 512GiB RAM	130,098.36 ops/s	1,273.873	Full disclosure report Executive summary Signatures Supplementary package
	300	Alibaba Cloud ecs.r8a.16xlarge 64×AMD EPYC 9T24 @ 3.7GHz vCPUs, 512GiB RAM	131,263.87 ops/s	1,285.285
	1000	Alibaba Cloud ecs.r8a.16xlarge 64×AMD EPYC 9T24 @ 3.7GHz vCPUs, 512GiB RAM	127,784.51 ops/s	1,251.216
System: AtlasGraph 3.0.0 Test sponsor: StarGraph Date: 2023-12-25 Queries implemented in: Rust System cost: 2,032,461 RMB (284,797.99 USD)	30	AWS r5d.12xlarge 48×Intel Xeon Platinum 8259CL vCPUs, 384GiB RAM	37,631.25 ops/s	132.133	Full disclosure report Executive summary Signatures Supplementary package
	100	AWS r5d.12xlarge 48×Intel Xeon Platinum 8259CL vCPUs, 384GiB RAM	48,764.08 ops/s	171.223
	300	AWS r5d.12xlarge 48×Intel Xeon Platinum 8259CL vCPUs, 384GiB RAM	48,311.63 ops/s	169.635
System: GraphScope Flex 0.23.0 Test sponsor: Alibaba DAMO Academy Date: 2023-07-13 Queries implemented in: C++ System cost: 99,236 USD	30	AWS r5d.12xlarge 48×Intel Xeon Platinum 8175M vCPUs, 384GiB RAM	33,180.87 ops/s	334.363	Full disclosure report Executive summary Signatures Supplementary package
	100	AWS r5d.12xlarge 48×Intel Xeon Platinum 8175M vCPUs, 384GiB RAM	33,625.36 ops/s	338.842
	300	AWS r5d.12xlarge 48×Intel Xeon Platinum 8175M vCPUs, 384GiB RAM	33,261.38 ops/s	335.175
System: TuGraph 3.3.4 Test sponsor: Ant Group Date: 2023-01-28 Queries implemented in: C++ System cost: 277,542 RMB (40,908.33 USD)	30	Alibaba Cloud ecs.g8y.16xlarge 64×Arm-based YiTian 710 vCPUs, 256GiB RAM	16,133.08 ops/s	394.372	Full disclosure report Executive summary Signatures Supplementary package
	100	Alibaba Cloud ecs.g8y.16xlarge 64×Arm-based YiTian 710 vCPUs, 256GiB RAM	16,966.26 ops/s	414.739
	300	Alibaba Cloud ecs.g8y.16xlarge 64×Arm-based YiTian 710 vCPUs, 256GiB RAM	13,532.62 ops/s	330.804
System: TuGraph 3.2.0 Test sponsor: Ant Group Date: 2022-08-16 Queries implemented in: C++ System cost: 291,176 USD	30	AWS r5d.12xlarge 48×Intel Xeon Platinum 8259CL vCPUs, 384GiB RAM	12,252.50 ops/s	42.079	Full disclosure report Executive summary Signatures Supplementary package
	100	AWS r5d.12xlarge 48×Intel Xeon Platinum 8259CL vCPUs, 384GiB RAM	12,934.61 ops/s	44.422
	300	AWS r5d.12xlarge 48×Intel Xeon Platinum 8259CL vCPUs, 384GiB RAM	12,721.24 ops/s	43.689
System: Galaxybase 3.3.0 Test sponsor: CreateLink Date: 2022-05-16 Queries implemented in: Java System cost: 263,282 USD	30	AWS r5d.12xlarge 48×Intel Xeon Platinum 8259CL vCPUs, 372GB RAM	9,285.86 ops/s	35.27	Full disclosure report Executive summary Signatures Supplementary package
	100	AWS r5d.12xlarge 48×Intel Xeon Platinum 8259CL vCPUs, 372GB RAM	8,501.21 ops/s	32.289
	300	AWS r5d.12xlarge 48×Intel Xeon Platinum 8259CL vCPUs, 372GB RAM	8,370.52 ops/s	31.793
System: TuGraph 1.1 Test sponsor: FMA Date: 2020-07-26 Queries implemented in: C++ System cost: 280,650 USD	30	AWS r5d.12xlarge 48×Intel Xeon Platinum 8175M vCPUs, 374GB RAM	5,436.47 ops/s	19.37	Full disclosure report Executive summary Signatures Supplementary package
	100	AWS r5d.12xlarge 48×Intel Xeon Platinum 8175M vCPUs, 374GB RAM	5,010.77 ops/s	17.854
	300	AWS r5d.12xlarge 48×Intel Xeon Platinum 8175M vCPUs, 374GB RAM	4,855.52 ops/s	17.301

Results as a CSV file

The audited LDBC SNB Interactive results displayed above are available as a CSV file.

LDBC-certified auditors

SNB Interactive audits can be commissioned from the following LDBC-certified auditors:

Márton Búr
Arnau Prat-Pérez
David Püroja

Legacy audited results

Social Network Benchmark Interactive, version 0.2.2

Future workloads

The LDBC SNB Interactive v2 workload is currently under development. See the TPCTC 2023 paper “The LDBC Social Network Benchmark Interactive Workload v2: A Transactional Graph Query Benchmark with Deep Delete Operations” and its slide deck for details.