5.1.1 About I/O Resource Management (IORM) in Exadata Database Machine

IORM provides many features for managing resource allocations. Each feature can be used independently or in conjunction with other features.

Storage is often shared by multiple types of workloads and databases, which can often lead to performance problems. For example, large parallel queries on a production data warehouse can impact the performance of critical OLTP in another database. You can mitigate these problems by over-provisioning the storage system, but this diminishes the cost-saving benefits of shared storage. You can also schedule non-critical tasks at off-peak hours, but this manual process is laborious.

IORM allows workloads and databases to share Oracle Exadata Storage Servers according to user-defined policies. To manage workloads within a database, you can define database resource plans, using Oracle Database Resource Manager which is enhanced to manage Oracle Exadata Storage Server I/O resources. You can also define a container database (CDB) resource plan that allows management for the pluggable databases (PDBs) that it contains. To manage multiple databases, you can define an interdatabase plan. Or, you can define a cluster plan to perform cluster-based resource management.

When there is contention for I/O resources, IORM schedules I/O by immediately issuing some I/O requests and queuing others. I/O requests are serviced immediately for workloads that are not exceeding their resource allocation, according to the resource plans. I/O requests are queued for workloads exceeding their resource allocation. Queued I/Os are serviced according to the priorities in the resource plans when the workload no longer exceeds its resource allocation. When the cell is operating below capacity and there is no contention for I/O resources, IORM does not queue I/O requests, and lets a workload exceed its resource allocation to consume available I/O resources.

For example, if a production database and test database share Oracle Exadata Storage Server resources, you can configure resource plans that give priority to the production database. In this case, whenever the test database load would affect the production database performance, IORM schedules the I/O requests such that the production database I/O performance is not impacted. This means that the test database I/O requests are queued until they can be issued without disturbing the production database I/O performance.

Flash IORM protects the latency of critical OLTP I/O requests in flash cache. When table scans are running on flash concurrently with OLTP I/O requests, the OLTP latency is impacted significantly. Flash IORM queues and throttles the table scan, and other low priority I/O requests. The critical OLTP I/O requests are never queued. When the flash disks are not busy serving critical OLTP I/O requests, the queued I/O requests are issued based on the resource allocations in the IORM plan.

• About the IORM Objective
  The I/O Resource Management (IORM) objective option specifies the optimization mode for IORM.
• IORM Plans
  Exadata I/O Resource Management (IORM) manages resources using various IORM plans.
• Resource Assignment Methods
  You can use shares or allocations to assign resources in an IORM plan.