How do I compare usable server storage, IOPS and throughput?
Storage capacity answers how much fits; IOPS, throughput and latency describe different aspects of how quickly data can be accessed. Advertised raw disk capacity may differ from the space available to applications. Compare the storage design against the actual read and write pattern.
Before you start
- Current data size, growth estimates and the proposed disk or volume specification.
Instructions
1. Calculate usable capacity
Account for any RAID layout, operating system, filesystem overhead, reserved space, logs, database indexes and temporary processing files. Check the units used for advertised and operating-system sizes. Include the working space required by imports, updates and recovery procedures instead of filling the disk to its nominal limit.
2. Match the performance measure
Small random database operations can depend on latency and IOPS; large sequential transfers can depend on throughput. Ask about sustained limits, shared resources and burst behaviour. A high advertised maximum does not establish the performance of your workload at its normal queue depth.
3. Verify growth and failure options
Confirm how storage can expand, whether shrinking is supported, and which changes require downtime. Identify the effect of a drive or volume failure and the separate recovery arrangements. Run only a bounded, authorised benchmark in a disposable test area; never benchmark by overwriting a live device.
Check the result
- Usable capacity covers the planned data and operational working space.
- Measured latency and throughput satisfy the application under a realistic workload.
Troubleshooting
- There is free space but database requests are slow
- Examine queueing, read/write latency and workload concurrency. Free capacity does not imply sufficient I/O performance.
Official references
Use the documentation for your installed version when a flag, service name or platform feature differs.