Disk Throughput Calculator
Calculate sequential or random disk throughput in MB/s from block size, IOPS, and link utilization.
Inputs
I/O request size, e.g. 4KB random, 128KB or 1MB sequential
Operations per second the device sustains at this block size
Percentage of theoretical throughput actually achieved on the interface
Effective Throughput
1,000.00MB/s
Theoretical Throughput
1,000.00MB/s
Effective Throughput
0.9766GB/s
Data Movable per Hour
3,515.63GB
Step by step
Values used
Block Size = 128 KB; IOPS = 8,000; Link/Interface Utilization = 100 %
Theoretical throughput
throughput = block_size × IOPS
Effective throughput
effective = theoretical × utilization%
Effective Throughput
= 1,000.00 MB/s
Theoretical Throughput
= 1,000.00 MB/s
Effective Throughput
= 0.9766 GB/s
Data Movable per Hour
= 3,515.63 GB
How it works
Throughput is the product of how much data moves per operation (block size) and how many operations complete per second (IOPS): a device doing 8,000 IOPS at a 128KB block size is moving roughly 1000MB/s. Real-world sustained throughput on a given link or interface rarely reaches the full theoretical figure because of protocol overhead, contention and controller limits, so an achievable link-utilization percentage is applied on top to produce a realistic effective figure.
Formulas
Theoretical throughput
throughput = block_size × IOPS
- B
- block size
- I
- IOPS
Effective throughput
effective = theoretical × utilization%
- u
- achievable utilization as a fraction
Frequently Asked Questions
Why do larger block sizes generally yield higher throughput?
Larger blocks amortize the fixed per-operation overhead (command issue, seek/latency, protocol framing) over more data, so sequential workloads using large blocks (128KB-1MB) typically report much higher MB/s than random 4KB workloads even on the same device, despite the device's IOPS ceiling actually being higher for small blocks.
Why is IOPS often higher for small blocks but throughput higher for large blocks?
IOPS and throughput trade off against each other: small blocks let a device complete more discrete operations per second (higher IOPS), but each operation moves less data, so total MB/s can be lower than a workload using fewer, larger operations that saturate the same underlying bandwidth.
What utilization percentage should I assume for planning?
For well-tuned sequential workloads on modern NVMe SSDs, 85-95% of theoretical throughput is common. Network-attached or shared storage, older SATA/SAS interfaces, and highly random workloads typically see larger gaps between theoretical and effective throughput, often 50-70%.