An application may be waiting on an EBS volume attached to EC2. Compare pending requests and completed read/write activity for the same AWS/EBS VolumeId over aligned 60-second periods.
Read the signals correctly before attributing a cause; queue length alone does not prove a performance shortfall. Also check how operation counts convert to IOPS.
Which metric/statistic combination is most appropriate?
Align VolumeId and periods; compare queue Average and completed Ops Sum, dividing counts by period seconds for IOPS.
Detailed explanation
Transferred bytes measure data volume, not waiting requests.
Transferred bytes measure data volume, not waiting requests.
Network traffic is not the pending-request count of the selected EBS volume.
Network traffic is not the pending-request count of the selected EBS volume.
This compares pending requests with completed counts; divide total operations by 60 to obtain average IOPS.
This compares pending requests with completed counts; divide total operations by 60 to obtain average IOPS.
A period's operation count must be divided by its seconds to obtain a per-second rate.
A period's operation count must be divided by its seconds to obtain a per-second rate.
Check official documentation
Compare the following points with the official documentation.
Review metric units/statistics, VolumeId, the Ops/s calculation, and I/O size and performance limits.
Official references
What to confirm
Calculate 100 IOPS from 6,000 operations in 60 seconds and explain why queue depth alone does not establish the cause.
Key points
- VolumeQueueLength
- VolumeReadOps
- VolumeWriteOps
Notes
- Environment: AWS SOA-C03 exam guide and the linked AWS documentation
- Reviewed against the linked AWS documentation on 2026-10-04.
Foundation review
Convert counts on an aligned time axis
Use VolumeQueueLength Average for pending-work trends. ReadOps/WriteOps Sum gives completed counts; divide by period seconds for a rate. A total of 6,000 operations in 60 seconds is 100 IOPS. Byte-based throughput has a different unit.
Use the queue as an investigation signal
Queue growth depends on demand and I/O size and alone does not prove a fault or insufficient IOPS. Correlate latency and transferred bytes with volume IOPS/throughput limits and EC2 EBS bandwidth. Avoid treating one queue-depth threshold as valid for every volume type.