Cache throttling

Use this page to view statistics for throttling applied to host write operations and RUN copy operations on the IBM® TS7700 Grid.

Throttling refers to the intentional slowing of data movement in order to balance or re-prioritize system resources in a busy TS7700. Throttling can be applied to host write and inbound copy operations.

Throttling of host write and inbound copy operations limits the amount of data movement into a cluster. This is typically done for one of two reasons:

The amount of unused cache space is low, or
The amount of data in cache that is queued for premigration has exceeded a threshold.

Host write operations can also be throttled when RUN copies are being used and it is determined that a throttle is needed to prevent pending RUN copies from changing to the immediate-deferred state. A throttle can be applied to a host write operation because the amount of unused cache space is low; because the amount of data in cache that needs to be premigrated is high; or, in the case of RUN copies, because an excessive amount of time is needed to complete an immediate copy.

The Cache Throttling graph displays the throttling that is applied to host write operations and to inbound RUN copy operations. The delay represents the time delay, in milliseconds, per 32 KiB of transferred post-compressed data.

Each numeric value represents the average of values for all active partitions in the cluster. Information shown includes:

Cluster: The name of the cluster affected.
Copy: The average delay, in milliseconds, applied to inbound copy activities.
Write: The average delay, in milliseconds, applied to host write operations, both locally and from remote clusters.

For information about the steps you can take to achieve peak network performance, refer to IBM® IBM TS7700 Series Best Practices - Understanding, Monitoring and Tuning the TS7700 Performance, WP101465, available from the TechDocs link in the Related information section.