Skip to content
Omnikalc

IOPS to throughput calculator: convert IOPS and block size to MB/s and latency

Enter the IOPS, the block size and the queue depth of a disk or a volume to get the throughput in MiB/s, the average latency by Little’s law and the time to move a given amount of data.

Updated on
Your inputs

Value out of range

KiB

Value out of range

Value out of range

GiB

Value out of range

Results

Throughput

78.1 MiB/s

Throughput in megabits per second
655 Mbps
Average latency (Little’s law)
1.600 ms
Normal for a SATA SSD or a busy NVMe
Time to move the data
1:49:14

IOPS tell how many operations a disk can do, and not how much data it moves. The IOPS to throughput calculator turns IOPS and block size into throughput in MiB/s, finds the average latency with Little’s law from the queue depth, and gives the time to move a given amount of data.

How the calculation works

  • Enter the IOPS of the drive or volume, the block size of each operation and the queue depth (the number of operations in flight).
  • The throughput is IOPS × block size, converted to MiB/s and to megabits per second.
  • By Little’s law, the average latency equals the queue depth divided by the IOPS.
  • Enter the amount of data to get the time to move it at this throughput.

Example

A drive does 20,000 IOPS with 4 KiB blocks and a queue depth of 32. The throughput is 20,000 × 4 / 1024 = 78.1 MiB/s (655 Mbps). The average latency is 32 / 20,000 = 1.600 ms. To move 500 GiB at this throughput takes 6,554 seconds, about 1 h 49 min. With larger blocks, the same IOPS would give a much higher throughput.

Why it matters

The calculator helps system administrators, database owners, video editors and people who shop for SSDs or cloud volumes to compare specifications that use different units. It also helps to check if a volume is the bottleneck of an application.

Good to know

The calculator uses the block size and the queue depth that you enter and assumes that the IOPS are sustained. Real devices vary with the workload (sequential or random, reads or writes), the temperature and the state of the drive. The latency is an average and not the 99th percentile that matters for many applications. MiB (1,048,576 bytes) are used here, and the megabit value converts them to 10⁶ bits.

The RAID capacity calculator gives the usable space of an array, the cloud storage cost calculator prices storage in the cloud, the home server cost calculator budgets a server, and the CCTV storage calculator sizes the disks for cameras.

Frequently asked questions

How are IOPS and throughput related?

Throughput is the number of operations per second times the size of each operation. With 4 KiB blocks, 20,000 IOPS make only 78 MiB/s, and with 1 MiB blocks, 1,000 IOPS make 1,000 MiB/s. This is why a drive can show a large number of IOPS and a small throughput on small random reads, and the reverse on large sequential reads. Always give the block size when you quote IOPS.

What is Little’s law for storage?

Little’s law says that the number of operations in flight equals the rate times the average time each one spends in the system: queue depth = IOPS × latency. If a drive handles 20,000 IOPS at a queue depth of 32, the average latency is 1.6 ms. It is useful to find the queue depth needed for a target IOPS: at 0.1 ms of latency, 100,000 IOPS need a depth of 10.

What are typical values for different devices?

A 7,200 rpm hard disk does 75 to 150 random IOPS, with a latency of 5 to 15 ms, and 100 to 250 MiB/s of sequential throughput. A SATA SSD does 50,000 to 100,000 IOPS and about 550 MiB/s, and an NVMe drive 500,000 to 1,000,000 IOPS and 3,000 to 7,000 MiB/s. Cloud volumes have their own caps for IOPS and throughput, and the lower one applies.

Why is the result lower than the datasheet?

Vendors quote the peak at a large queue depth, with the best block size, on a fresh drive. In real use, mixed reads and writes, small queue depths, the file system, the encryption and the wear of the drive reduce the numbers. Write operations are often slower than reads, and a sustained write can fall when the cache is full. Measure with a tool such as fio for your own workload.

Was this calculator helpful?

Thanks for your feedback!

Results are estimates provided for information purposes. They do not replace professional advice.