Oracle Database QoS Management measures use and wait times for CPU, Global Cache, I/O, and other resources to determine where a bottleneck is located.

The target Performance Class and its bottle-necked resource are identified on the Oracle Database QoS Management Dashboard (the Dashboard), however, only the CPU resource is actively managed in this release.

A CPU resource bottleneck is detected when there are excessive wait times on the collection of CPU queues running that workload. Oracle Database QoS Management offers recommendations you can implement to relieve the bottleneck.

One solution to this type of bottleneck is to increase the number of opportunities for the workload to run on the CPU. Oracle Database QoS Management implements this solution by assigning the workload to a consumer group that has more CPU shares across the server pool.

Another solution is to provide more CPU resources. If you have multiple instances sharing the CPU resources for each server in the server pool, and you have implemented instance caging, then Oracle Database QoS Management can suggest altering the CPU counts for the instances in the server pool; this solution gives more CPU resources to the workloads that are not meeting performance expectations by taking CPU resources away from an instance that is of lower rank or has the headroom to contribute the resources.

If there is a CPU resource bottleneck that cannot be relieved by adjusting the CPU counts between instances, then Oracle Database QoS Management can recommend moving a new server into the server pool. The server can come from the Free pool, from a less-stressed server pool, or from a server pool that hosts a less critical workload.

A Global Cache resource bottleneck is detected when there is excessive data block movement between database instances. This is usually caused by an application that is not configured properly or is not able to scale horizontally. Configuring the application to run in a server pool with a maximum size of one or partitioning the data can usually relieve the bottleneck.

Oracle Database QoS Management cannot perform either of these actions in this release and does not provide a recommendation that can be implemented for this type of bottleneck.

An I/O resource bottleneck is detected when there are excessive wait times on the storage subsystem. This type of bottleneck is typically caused by either too few disk spindles or not enough network bandwidth on the storage interconnect. To resolve this bottleneck, spread the database files across a higher number of disks, or configure a separate network interface card (NIC) for a dedicated storage interconnect.

Oracle Database QoS Management cannot resolve this type of bottleneck in this release and does not provide a recommendation that can be implemented.

The last resource type used to categorize bottlenecks, Other, is used for all other wait times. These database wait times are usually caused by SQL performance issues that result from an application that is not optimized, waiting on latches, and so on. These bottlenecks can be investigated using Oracle Database tuning tools such as Automatic Workload Repository (AWR) and Automatic Database Diagnostic Monitor (ADDM).

Resolving these types of bottlenecks are outside the scope of the run-time system management provided by Oracle Database QoS Management and Oracle Database QoS Management does not provide any recommendations that can be implemented