Three SQL Server pods that are configured with 32 CPUs per pod were deployed across three worker nodes (one pod per node) to maximize the compute capacity of the worker nodes. Three HammerDB instances were deployed to generate the OLTP load simultaneously by running the benchmark. One to one mapping is considered between the HammerDB instance and SQL pods to simulate a production scenario.
The following graph shows the TPM performance for the three SQL Server pods running on the Validated Platform:
As shown in the above test results, 3.3 million+ TPMs were achieved with sub millisecond read and write latency.