5.1.1.1 About the IORM Objective

The I/O Resource Management (IORM) objective option specifies the optimization mode for IORM.

The IORM objective is configured individually on every storage server using the CellCLI ALTER IORMPLAN command. To deliver consistent overall system performance, ensure every storage server in the storage cluster uses the same IORM configuration settings.

The valid objective values are:

• auto - Causes IORM to determine the best mode based on active workloads and resource plans. IORM continuously and dynamically determines the optimization objective based on the observed workloads and enabled resource plans. For most use cases, auto is the recommended value.

  Starting with Oracle Exadata System Software release 21.2.0, auto is the default setting for new deployments.

• high_throughput - Optimizes critical DSS workloads that require high throughput. This setting improves disk throughput at the cost of I/O latency.

• low_latency - Optimizes critical OLTP workloads that require extremely good disk latency. This setting provides the lowest possible latency at the cost of throughput by limiting disk utilization.

• balanced - Balances low disk latency and high throughput, which is useful for a mixture of critical OLTP and DSS workloads. This setting limits disk utilization of large I/Os to a lesser extent than low_latency to achieve a balance between latency and throughput.

• basic - Use this setting to limit the maximum small I/O latency and otherwise disable I/O prioritization. This is the default setting for new deployments in Oracle Exadata System Software release 20.1.0 and earlier. Specifically, using this setting:

  • IORM guards against extremely high latencies for log writes, buffer cache reads, and other critical I/Os by prioritizing them ahead of smart scans and other disk intensive I/O operations.
  • IORM manages access to flash cache and flash log.
  • IORM manages scans to ensure that resources are shared between different workloads.
  • IORM does not enforce the maximum utilization limits in the IORM plan. Plan allocations are only used to guard against extremely high latencies, and there is no consideration for plan conformance. For stricter plan conformance, and enforcement of maximum utilization limits, the objective option must be set to something other than basic.

On High Capacity (HC) storage servers, flash IORM protects OLTP latency by prioritizing critical I/Os to flash devices ahead of smart scan and low priority I/Os. This occurs by default and regardless of the objective option setting.

On Extreme Flash (EF) storage servers, the IORM objective provides some additional control over flash IORM. On EF storage servers, flash IORM behaves the same as on HC servers when the objective option is set to basic, auto, or balanced. However, on EF storage servers, high_throughput increases the scan throughput at the cost of critical I/O latency, and low_latency provides maximum latency protection at the cost of significant scan throughput degradation when both workloads run concurrently.