How do I know if CPU is causing a performance issue?

ONTAP 9

Performance is about expectations
- Knowing if you have a problem requires understanding your expectations
Expectations can be implicit or explicit
- Implicit expectations – AFF platforms should see latency < 2ms (on average)
- Explicit expectations – MS Exchange requires latency to average < 20ms
Are users filing tickets/reporting issues?
- Users filing support tickets or raising concerns about performance for the storage environment
- Are applications generating alarms regarding performance?
- Are jobs failing or taking longer than expected?

Using ActiveIQ Unified Manager, drill down into the Volume Performance
- Looking at the Cluster Components, the two main CPU resources are:
  - Network Processing
  - Data Processing
- If these two components make up the majority of your latency, then CPU is the primary factor in overall latency.
- Note: Latency coming from CPU is normal. In most AFF systems, CPU is the primary contributor to latency. It would be a 'problem' only if the latency seen is higher than expected.
Using CLI to collect QoS statistics from a cluster in real-time
- This will show you the contributions to a workload’s latency by delay center
  - Network and Data are the two CPU resources in this display
- As with UM, if these two components comprise the majority of latency, then CPU is the primary factor in overall latency