Monday, March 9, 2015

Enabling Predictive Cache Statistics (PCS) for Data OnTap 8.2p

* node1 is the controller currently primary to the aggregate/vol/LUN.

Step 1: Enable PCS
node1::> node run –node node1
node1::> options flexscale.enable on
node1::>options flexscale.enable
flexscale.enable pcs  you should see this
node1::>options flexscale.pcs_size 330GB  based on 3 x 200GB SSD RAID4

Step 2: Allow the representative workload to run and Run your workload

Step 3: Collect data throughout the process
node1::>stats show -p flexscale-access
NetApp recommends issuing this command through an SSH connection and logging the output throughout the observation period because you want to capture and observe the peak performance of the system and the cache. This output can also be easily imported into spreadsheet software, graphed, and so on. This process initially provides information on the “cold” state of the emulated cache. That is, no data is in the cache at the start of the test, and the cache is filled as the workload runs. The best time to observe the emulated cache is once it is filled, or “warmed”, as this will be the point when it enters a steady state. Filling the emulated cache can take a considerable amount of time and depends greatly on the workload. References:
  1. Introduction to Predictive Cache Statistics
  2. Clustered_Data_ONTAP_82_System_Administration

No comments: