IOPS to Throughput Calculator

About the IOPS to Throughput Calculator

IOPS and throughput are two sides of the same storage performance coin, connected by block size. A device capable of 100,000 random 4K IOPS delivers about 390 MB/s of throughput. But that same device doing 256K sequential IOs at 10,000 IOPS delivers 2,500 MB/s. Understanding this relationship is essential for matching storage to workload requirements.

Database workloads typically use small random IOs (4K–16K) where IOPS is the bottleneck. Media streaming and backup workloads use large sequential IOs (128K–1M) where throughput matters more. This calculator converts between IOPS and throughput for any block size, helping you translate vendor specs into real-world performance expectations for your specific workload.

When This Page Helps

Storage vendors often quote peak IOPS and peak throughput separately, but you can't achieve both simultaneously. This calculator shows the actual throughput for your IO pattern's block size, bridging the gap between specs and real-world performance.

How to Use the Inputs

1. Enter the IOPS (IO operations per second).
2. Enter the block size in KB.
3. Review the calculated throughput in MB/s and GB/s.
4. Try different block sizes to see how throughput scales.
5. Compare against your workload's typical block size.

Example Calculation

Tips & Best Practices

• Databases (OLTP) typically use 4K–8K random I/O; optimize for IOPS.
• Streaming and backup workloads use 128K–1M sequential I/O; optimize for throughput.
• NVMe drives can sustain high values for both IOPS and throughput simultaneously.
• Cloud storage often bills per IOPS provisioned—right-size block size to avoid overpaying.
• RAID write penalties reduce effective write IOPS: RAID 5 penalty is ~4×, RAID 6 is ~6×.
• Queue depth affects achievable IOPS; typical databases need queue depth 8–32 for peak performance.

Understanding the IOPS-Throughput Relationship

Think of IOPS as the number of trips a delivery truck makes, and throughput as the total cargo delivered. Small packages (4K blocks) mean many trips but limited cargo per trip. Large packages (1M blocks) mean fewer trips but much more cargo per trip. The truck (storage controller) has finite speed and capacity.

Cloud Storage Implications

Cloud providers often charge per provisioned IOPS. Understanding the IOPS-to-throughput relationship helps you avoid over-provisioning. If your workload is sequential with large blocks, you may need fewer IOPS than expected to achieve the required throughput.

Benchmarking Tips

When benchmarking storage, always test with the block size and access pattern that matches your production workload. A vendor's advertised 1 million IOPS at 512 bytes may translate to only 3,906 MB/s—and your database needs 8K random reads, not 512-byte operations.

Frequently Asked Questions

• What block size should I use?

  Use the block size that matches your workload. Databases typically use 4K–8K for random operations and 64K–256K for sequential scans. File servers use 4K–64K. Media streaming uses 128K–1M. Check your application's IO pattern with tools like iostat or perfmon.

• Can a drive deliver max IOPS and max throughput at the same time?

  No. Max IOPS is achieved with small blocks (4K) and max throughput with large blocks (128K+). The drive's controller has a fixed processing bandwidth, so there's always a trade-off between the number of operations and the size of each operation.

• How does this apply to cloud storage?

  Cloud block storage (EBS, Azure Disk, Persistent Disk) specifies both IOPS and throughput limits. Your volume hits whichever limit first. If you provision 3,000 IOPS and 125 MB/s, 4K random IO is IOPS-limited at 11.7 MB/s, while 256K IO is throughput-limited at 500 IOPS.

• Why do SSD specs show different IOPS for different block sizes?

  Smaller blocks require less data transfer per operation, so the controller can process more operations per second. As block size grows, data transfer becomes the bottleneck. A drive might do 500K IOPS at 4K but only 5K IOPS at 1M because it's saturating its throughput limit.

• What is the typical block size for databases?

  Most databases use 4K–8K pages for random access (index lookups, row fetches) and larger blocks (64K–256K) for sequential operations (table scans, backups). Oracle uses 8K default, PostgreSQL uses 8K, SQL Server uses 8K–64K.

• How do mixed workloads affect this calculation?

  For mixed workloads, calculate IOPS and throughput separately for each IO pattern, then sum them. Ensure both the total IOPS and total throughput stay within your storage device's limits. Most sizing tools model this as a weighted average.