Performance Read Cache Management Tasks - IBM System Storage DS3000 Programming Manual

v Average response time – The performance modeling tool uses calculated response times from the
operation of the Performance Read Cache to estimate the time it takes to run during each external and
internal I/O operation. The tool uses these response time measurements along with measurements of
I/O operations performed during the run of a workload that is applied to the base volumes with
Performance Read Cache enabled. Keep in mind that workload on other volumes can affect these
results due to increased response time. The resulting estimates are of the average response time for
external I/O operations. When making a decision to increase or decrease the capacity of your current
Performance Read Cache, look for an acceptable response time, and then compare that with the
response time of your current Performance Read Cache capacity.
You can make changes by using options outside of the performance modeling tool. For example, you can
change the I/O characteristic types and enable or disable Performance Read Cache on base volumes.
These two parameters control the working set for the performance modeling. You change the I/O
characteristic types by selecting these menu options: Storage >> Performance Read Cache >> Change
I/O Type. You enable or disable Performance Read Cache on base volumes by selecting these menu
options: Storage >> Volume >> Performance Read Cache.
Thrashing
Thrashing is a condition where data is constantly being read from the base volume into the Performance
Read Cache, but not accessed again before it is cleared from the Performance Read Cache. The data is
cleared from the SSD cache because all Performance Read Cache blocks are allocated and another volume
or logical block addressing (LBA) range needs to allocate a cache block that currently had data. The
Performance Read Cache processes do not move data from the base volume into the Performance Read
Cache if only one read occurs into a cache block; therefore, a purely random read workload does not
cause thrashing and unnecessary overhead with any working set size.
To determine whether cache thrashing might be occurring, look at the Recycle Actions statistic in the
detailed statistics that you can view when you save the Performance Read Cache statistics to a ..csv file

Performance Read Cache management tasks

As soon as you create the Performance Read Cache all of the volumes assigned to the Performance Read
Cache can start using it. The Performance Read Cache is a non-addressable volume that requires little
maintenance or management. You can, however, perform these tasks to make sure that the Performance
Read Cache runs best to support your storage subsystem:
v View information about the drives, status, and capacity of the Performance Read Cache.
v Locate the drives that physically comprise the Performance Read Cache.
v Adding drives to and removing drives from the Performance Read Cache.
v Suspend and resume Performance Read Cache operation.
v Rename the Performance Read Cache.
Viewing information about the Performance Read Cache
When you create your Performance Read Cache, you provide a name and select the solid state disk (SSD)
drives for the cache. However, the firmware maintains more information about the Performance Read
Cache, such as maximum capacity, consumed capacity, size of the drives, and other information. You can
see all of the information about the Performance Read Cache by running this command:
show performanceReadCache [performanceReadCacheName]
This command returns information about the Performance Read Cache that is similar to this example.
Performance Read Cache name: my_cache
Status:
Type:
I/O characteristic type:
Optimal
Read Only
File System
Chapter 10. Using the Performance Read Cache premium feature
10-5