Learn the measurements
IOPS counts operations while throughput measures transferred data over time. Many small reads differ from streaming large files. I would describe the request size and pattern before comparing volume names, because one performance number cannot explain every workload.
Inspect the whole path
The instance and application can limit performance as well as storage. Increasing volume performance may not fix a CPU bottleneck or inefficient query. Check the relevant instance capability and measure representative work instead of testing an unrelated pattern.
Change one thing deliberately
Choose a suitable starting option with current documentation and pricing. Record the result, adjust a justified setting and compare again. Keep cost in the experiment so that a faster result can be evaluated against what it costs.
Try it yourself
Compare a boot disk, transactional dataset and sequential processing job. Identify which storage measurements matter most for each.
A common question
Are IOPS and throughput interchangeable?
No. They describe different aspects of storage work; request size and access pattern connect them in practice.
References & video notes
Use the video to follow the topic with me. Console screens and older offers can change, so check the current AWS documentation before performing a lab.
Keep learning
- Block, file and object storage: choose the access pattern
- Instance store vs EBS: where should your data live?
- EBS snapshots: include a restore in your backup lab