The LDBC Social Network Benchmark Interactive workload v2: A transactional graph query benchmark with deep delete operations

TL;DR

This paper introduces the Interactive v2 benchmark for transactional graph databases, featuring delete operations, advanced queries, support for larger datasets, and a new temporal parameter algorithm to ensure consistent performance.

Contribution

It presents a renewed version of the industry-standard social network benchmark with new features and improved stability for graph query performance.

Findings

Includes delete operations and complex path queries.

Supports larger datasets for scalability testing.

Introduces a temporal parameter curation algorithm.

Abstract

The LDBC Social Network Benchmark's Interactive workload captures an OLTP scenario operating on a correlated social network graph. It consists of complex graph queries executed concurrently with a stream of updates operation. Since its initial release in 2015, the Interactive workload has become the de facto industry standard for benchmarking transactional graph data management systems. As graph systems have matured and the community's understanding of graph processing features has evolved, we initiated the renewal of this benchmark. This paper describes the draft Interactive v2 workload with several new features: delete operations, a cheapest path-finding query, support for larger data sets, and a novel temporal parameter curation algorithm that ensures stable runtimes for path queries.