Target utilization in an Oracle RAC active/active configuration

When deploying an active/active failover configuration like Oracle RAC, take into consideration your target node utilization, especially during the seasonal peak periods. Although RAC is an active/active configuration, you have to be careful when you run the RAC nodes at close to max utilization. First, RAC needs CPU to maintain the Cache Fusion. Secondly, during failover, all the work from the failed node eventually spills over to the remaining node. For example, at the peak hour, the combination of high volume agent processing and application server processing is driving a 4-node RAC configuration to 80% busy. If a node fails, the servers reconnect to the remaining node. The combined workloads drive the database to a point where performance degrade or at worst the system becomes unstable.

Ideally, you should try to keep the average node utilization below 60% utilization (for a 4-node RAC) to reduce the risk of overloading the remaining nodes. In the event of a node failure, the remaining node will likely grow to 80%.

Alternatively, if you typically run the nodes at higher utilization, you should consider classifying workloads into priority groups. For example, transactions that are initiated by customers may be put into the high priority group because they add to the overall customer experience. These could include application server workloads. Customers with short “click to release” service levels may consider agents that schedule and release orders as high priority workloads. Workloads such as order purge or other maintenance work could be categorized into lower priority groups.