HammerDB is a free, open-source, database load-testing and benchmarking tool that works with databases like Oracle MySQL, SQL Server, and others. With HammerDB you can create a test schema and load it with test data. After the database is loaded with test data, you can run both transaction and analytic scenarios on your Dell EMC infrastructure. You can use the performance metrics from the load tests to characterize the performance of the system and then to test changes to further improve performance.
For the validation tests in this paper, we used HammerDB to run an OLTP-style workload. OLTP workloads consist of many short duration database transactions that perform highly selective lookup, modify, create, or delete operations. Online retail is a widely used example of an OLTP workload that has many characteristics in common with custom enterprise applications that are developed to support internal and external businesses processes.
For test purposes, our goal was to generate a load consistent with an average enterprise database. We tested small to medium-sized production and development workloads without attempting to stress-test the infrastructure to the maximum limit. Dell Technologies quality assurance engineers want to validate that the application behaves as intended during typical MySQL InnoDB Cluster operations. Our testing goal was to demonstrate how customers might use an entry-level infrastructure footprint for MySQL databases in a functional production and development environment.
To scale the HammerDB TPC-C data generation, we specified the number of simulated warehouse facilities from which a fictitious company can supply products to fulfil orders from fictitious customers. We used a scale factor of 15,000 warehouses when generating data for the TPC-C schema that resulted in approximately 1.5 terabytes of data without compression or deduplication. For additional clarity, note that the term ‘warehouse’ in the context of a HammerDB TPC-C test has no relationship to any Data Warehousing workloads or decision support systems. TPC-C defines a transaction-based system and not a decision support system (DSS).
In loading the 1.5 terabytes of data, Dell Technologies engineers found that the PowerStore T performed well and without issues. After the data load was complete, the engineering team ran the TPC-C workload without issue. No performance measurements will be communicated because the goal of these tests was to validate the PowerStore T as a target platform for MySQL databases.