Performance Monitoring and Clustered Data ONTAP

With a snap of the fingers I have been at NetApp for almost 6 months already! I have been quite busy acclimating to life as a consultant as I have been in the field almost 100% of the time since May. While being busy has been nice, it has left me with little time to update this blog. Fortunately, I have finally discovered some late night motivation (caffeine) and will be publishing several articles in the coming days regarding topics which I commonly deal with in the field.

With each new Clustered Data ONTAP (cDOT) deployment I have been involved with, each customer seems to ask me the same question: “How do I perform a root cause analysis of a possible performance issue in regards to our new cDOT system(s)?”. In typical NetApp fashion we offer several straightforward and easy to use tools to help fulfill customer needs regarding performance. In the list below I have ordered the tools in order in which they should be utilized to troubleshoot a possible performance issue:

  • OnCommand Performance Manager
    • OnCommand Performance Manager (OPM) is a relatively new product that was released shortly after 8.2 went GA. OPM was built from the ground up and allows administrators to monitor their cDOT environment(s) for performance related issues. Unlike Performance Advisor, OPM has a greatly simplified interface with just a few controls. This design choice is intentional as the philosophy for how performance monitoring is achieved has evolved with the release of cDOT. Rather than present an interface with hundreds of metrics, OPM displays the relevant metrics ONLY when it detects a performance-related event. The simplicity of the interface allows administrators to more efficiently perform root cause analysis tasks and thus free up more of their time for other tasks.
  • CLI Commands
    • On top of leveraging OPM, you can also rely on several CLI commands listed below which display basic performance data in realtime:
      • # statistics show-periodic -interval 1
        • This command by default will show cluster-wide statistics however if you want to see node-specific data you can append “-node nodename” to the end of the command. The output of this command is very comparable to that of the 7-mode “sysstat” command.
      • # cluster statistics show
      • # dashboard performance show
    • While the last two commands produce more simplistic output than that of the “statistics show-periodic” command, they can still be useful depending on the data you are looking for.
  • QoS Approach
    • With the release of cDOT 8.2, QoS functionality was introduced as a way to control workloads and thus ensure that performance is never compromised. As described by a fellow NetApp Community contributor QoS can also be utilized to examine performance of a given workload by assigning storage objects to an “INF” QoS policy. Once the appropriate storage objects are assigned to the “INF” QoS policy, you can execute the following commands to monitor its performance:
      • # qos statistics workload performance show -refresh-display true
      • # qos statistics performance show
      • # qos statistics characteristics show
      • # qos statistics latency show

For more details please refer to the following links: