We used SLOB 2.4 to run the OLTP test case. This test case demonstrates a realistic workload in which customers access the most recent data, which is a small portion of the whole database. The data is partially cached in PowerMax and 60% of the data reads are serviced from the PowerMax cache. We focused on achieving high IOPS while maintaining relatively low latencies.
The figure below shows the Oracle AWR IOPS for this test case. On average, the database performance during the test was 183,300 read IOPS and 47,258 write IOPS, for a total of 230,558 IOPS.
Figure 10. AWR IOPS results
The figure below shows the Oracle AWR response times for the OLTP test case. On average, the data files read response time was 406.5 microseconds (0.40 ms), and the redo log writer response time was 727.40 microseconds (0.74 ms).
Figure 11. AWR response time results
The figure below shows Unisphere performance metrics for Oracle redo log files (redo_sg storage group). The test produced 860 IOPS with a consistent 0.5 ms write response time.
It is evident that with a realistic workload, our lab environment of single-brick PowerMax storage arrays was able to consistently generate close to a quarter million IOPS with sub millisecond read and write latencies. To achieve very low latencies, we need to keep the system utilization moderate, but not at maximum (due to I/O queueing, as IOPS increase, latencies may also increase). We can achieve higher IOPS and maintain low latencies by scaling up the system, such as adding PowerMax bricks, iSCSI front-end ports, or database servers.