Why would AIQUM report a higher delay than other performance monitoring tools?
Applies to
- ONTAP 9
- ActiveIQ Unified Manager (AIQUM)
- ActiveIQ PAS (AIQ PAS)
Answer
AIQUM has a differenty way to report the latency than the other performance monitoring tools. This will lead to a significant higher delay from a Service/Delay center in AIQ PAS if there is any unaccounted latency.
AIQUM:
- It calculates the raito of different Service/Delay centers with the total calculated latency
- Then it use that ratio to display the contribution of a Service/Delay center to the actual total measured latency
The rest tools (CLI/AIQ PAS/Perfstat):
All these tools will display the actual measured latency
Example:
- Out of the 58ms total latency, only 772us is properly accounted.
- Only 487us is accounted under Cluster (Cluster Interconnect)
- But, AIQUM will report the 58ms total latency and 36.5ms delay from Cluster.
- The ratio of the delay from Cluster against the total latency is 487.00us / 772.00us ~= 63%
- The calculated delay from Cluster is 58ms x 63% ~= 36.5ms
> qos statistics volume latency show
Workload ID Latency Network Cluster Data Disk QoS Max QoS Min NVRAM
--------------- ------ ---------- ---------- ---------- ---------- ---------- ---------- ---------- ----------
test_volume111 41567 772.00us 188.00us 487.00us 67.00us 0ms 0ms 0ms 30.00us
> qos statistics volume performance show
Workload ID IOPS Throughput Latency
--------------- ------ -------- ---------------- ----------
test_volume111 41567 17 86.74KB/s 58.00ms
Notes:
- AIQUM displays a cooked number for each Delay/Service center
- When there is no gap between the total latency and the sum of all the Delay/Service centers, what AIQUM displays agrees with the other tools
- When there is a significant delay between the total latency and the sum of all the Delay/Service centers, what AIQUM displayes could be distorted