Silicon Graphics InfiniteStorage 4000 Series, Настройка и обслуживание

Страница: 185 / 276

Chapter 10: Using the SSD Cache Premium Feature 173
Creating the
SSD Cache,
Adding
Volumes, and
Removing
Volumes Before you create the SSD cache, make sure that suitable SSD drives are available on
the storage array. You can achieve the best performance when the working set of the
data fits in the SSD cache so that most host reads can be serviced from the lower
latency solid state disks instead of the higher latency hard drives (HDDs).
To create the SSD cache run this command:
create ssdCache
userLabel=" ssdCacheName "drives=(trayID1,drawerID1,slot
ID1 ...
trayIDn,drawerIDn,slotIDn)[updateExistingVolumes=(TRUE
|FALSE)]
To use this command you need only to create a name for the SSD cache and identify
the specific SSD drives that you want to include in the cache. After you run the
command, all of the volumes in the storage array can use the SSD cache.
If you do not want all of the volumes in the storage array to use the SSD cache, set the
updateExistingVolumes
parameter to
FALSE
. When you create the SSD
cache without volumes, you can later assign specific volumes to the SSD cache by
running this command for each volume that you want to add:
set volume [" volumeName "] ssdCacheEnabled=TRUE
You can assign standard volumes, snapshot volumes, or consistency group snapshot
volumes. You can assign only volume at a time to the SSD cache.
If you do not want a specific volume to use the SSD cache, you can remove the
volume from the SSD cache by running this command:
set volume [" volumeName "] ssdCacheEnabled=FALSE
After you have created the SSD cache, you can model the performance of the SSD
cache to make sure that the SSD cache is running as required for you storage array.
SSD Cache
Performance
Modeling After you create the SSD cache, you can run a performance modeling tool to help
determine the best cache size for your storage array.
The performance modeling tool helps you determine the performance improvement
for SSD cache capacity by running a workload on the SSD cache that has the same
characteristics as the workload that you run in production. Performance modeling
monitors and measures I/O activity for a period of time and estimates performance for
various SSD cache sizes. The performance modeling tool then shows the actual
performance for the SSD cache that you created.
Depending on the cache capacity and workload, it might take 10 to 20 hours to fully
populate the cache. The cache has valid information even after a run of a few minutes,
but the longer that you can run the performance modeling tool the more accurate the
results of the metric data.