Oracle 10g: user scaling

Edit online

To see how the system reacts to changing workloads the amount of users is scaled from 80 to 120 in steps of 20 users.

Normalized transactional throughput and CPU load when scaling the amount of workload generating users

The result for throughput and CPU load is shown in Figure 1.
Figure 1. Normalized transactional throughput and CPU load when scaling the amount of workload generating users

graph shows normalized transactional throughput and CPU load when scaling the amount of workload generating users

Observation

The transactional throughput and the CPU load scale in a very similar fashion. But while the amount of users is finally increased by 50%, the transactional throughput only increases by 18%.

Conclusion

The fact that the CPU load and transactional throughput scale at the same rate, indicate that there is no overhead related to the workload in the range. The fact that both scale much lower than the amount of users, shows that the workload is waiting for disk I/O.

Disk I/O throughput when scaling the amount of workload generating users

Figure 2 shows the disk throughput rate when the amount of users is scaled.

Figure 2. Disk I/O throughput when scaling the amount of workload generating users. The right axis show the percentage relative to 80 users.

Graph shows disk I/O throughput when scaling the amount of workload generating users. The right axis show the percentage relative to 80 users.

Observation

The disk I/O throughput rate scales in the same manner as the transaction rates. The disk transfer rates exceed 200 MiB/sec. The dominant wait event class is USER I/O which covers nearly 100% of the wait times, it increases when the amount of users is scaled from 80 to 120 by 50%.

Conclusion

The disk transfer rates of 200 MiB/sec could be considered as high for a database workload. This shows clearly that the test system is suitable for testing variations in storage connectivity setup, especially in combination with the high role of the wait event class USER I/O. It also shows that the used storage server and the attachment provide a relatively high disk I/O bandwidth and are no bottlenecks.