HammerDB

HammerDB is a popular open-source database benchmarking and load-testing tool that runs standardized, TPC-derived workloads against many relational databases, including Oracle, SQL Server, MySQL, PostgreSQL, Db2, and MariaDB. It lets practitioners reproduce TPC-C-style OLTP and TPC-H-style analytics on their own hardware without the cost and rules of an official, audited TPC submission.

What It Measures

HammerDB implements two derived workloads. TPROC-C is a TPC-C-like OLTP test reporting NOPM (new orders per minute) and TPM (transactions per minute). TPROC-H is a TPC-H-like decision-support test reporting query execution times across the analytical query set. These metrics characterize transactional throughput and analytical query performance respectively. HammerDB deliberately renames the workloads to make clear they are not official, audited TPC results.

Methodology

For TPROC-C, HammerDB builds the warehouse schema at a chosen number of warehouses, then drives a configurable number of virtual users that execute the standard transaction mix using stored procedures for efficiency. You set warehouse count, virtual-user count, ramp-up and test durations, and keying/think-time options. The tool collects NOPM and TPM over the measurement interval and can run an automated series across user counts to find the throughput curve. TPROC-H generates data at a scale factor and runs the analytical query stream, recording per-query and total times. HammerDB provides both a GUI and a command-line/CLI scripting interface for repeatable, automated runs.

How to Interpret Results

Report warehouse count (or scale factor), virtual-user count, and whether keying and think times were enabled — disabling them inflates throughput dramatically and is common in informal runs. Use the automated user sweep to identify peak NOPM and the point where latency degrades, rather than a single number. Because these are derived, unaudited workloads, NOPM is not directly comparable to official tpmC. Compare engines only under identical HammerDB configurations and hardware.

Limitations

HammerDB results are not official TPC numbers and must not be presented as such; the schema and queries are derived approximations. Results are highly sensitive to settings, especially think times and connection pooling, which makes published community numbers hard to compare. It focuses on relational OLTP and analytics and does not cover NoSQL or document stores. Use it as a practical, reproducible cross-engine load test under carefully documented conditions. Its broad engine support and scriptable CLI make HammerDB the most practical way for teams to reproduce TPC-style workloads in house and track performance across database versions over time. Used carefully with documented think times and a full user sweep, it gives a defensible relative picture even though its numbers carry no official TPC standing.