Home > Storage > PowerFlex > White Papers > Dell APEX Block Storage for Azure: IBM Db2 Database Deployment Best Practices > TPROC-C workload using HammerDB
The HammerDB TPROC-C workload is the OLTP workload derived from the TPC-C Benchmark Standard with modifications to make running HammerDB straightforward and cost-effective on any of the supported database environments. This test starts with one Azure Red Hat Linux VM running an IBM Db2 database and perform TPROC-C workload on a single Db2 database. Later increase the number of Azure Red Hat Linux VMs running the Db2 database and perform the TPROC-C workload in parallel.
The Db2 database scalability test is used to monitor the Db2 database performance as the number of Db2 databases is increased, where each database is running on a separate Azure Red Hat Linux VM. The Azure Standard F16s v2 VM with 16 vCPUs and 32 GB memory is used for installing and configure Db2 database. The Db2 database created with an 8 K block size and 4 GB buffer pool. HammerDB is used to create a 1,000 warehouse Db2 TPROC-C schema and populate approximately 90 to 100 GB test data for the workload test.
The test process included the following steps:
The following table shows the results of the TPROC-C workload with the scaling of the Db2 databases on APEX Block Storage for Azure environment.
Metrics | 1 x Db2 database | 2 x Db2 databases | 3 x Db2 databases | 4 x Db2 databases |
Transactions per minute (TPM) | 119,546 | 209,061 | 328,200 | 476,521 |
Read IOPS | 11,233 | 20,300 | 31,914 | 38,795 |
Write IOPS | 5,843 | 10,455 | 16,726 | 23,020 |
Total Read + Write IOPS | 17,076 | 30,755 | 48,640 | 61,815 |
Db2 data file read latency (ms) | 1.01 | 1.01 | 1.00 | 1.02 |
Note: The Db2 database performance tests are limited by the available database license which only allowed for 8 CPU cores to be engaged in each VM. OLTP database performance tests have high dependency on available CPU cores. It is assumed that when running the Db2 database with a less restrictive license the performance numbers will closely match the FIO numbers as far as IOPS and bandwidth per VM.
The following figure shows that the Db2 database transactions per minute scaled linearly with the number of Db2 databases, while the average Db2 database files read response time remained constant at 1 millisecond even when four Db2 databases were deployed on APEX Block Storage for Azure environment.