IOPS is a performance metric used to characterize computer storage devices like HDD, SSD, and SAN. It measures the maximum number of read and write operations to storage locations per second. While throughput measures the volume of data transferred (MB/s), IOPS focuses on the frequency of discrete transactions, which is vital for assessing system responsiveness.
The metric is influenced by the data block size (typically 4 KB), the read/write mix, and the access pattern (random vs. sequential). In SSDs, IOPS is driven by controller efficiency and NAND flash parallelism. In HDDs, it is physically constrained by rotational speed (RPM) and the time required for the actuator arm to seek specific sectors.
High IOPS is essential for workloads involving frequent, small-sized data transfers, such as relational databases (SQL), virtual desktop infrastructure (VDI), and real-time analytical processing. Low IOPS in these environments leads to increased I/O wait times and application latency.
How it works
A standard 7200 RPM SATA drive provides roughly 80 IOPS for random access. In contrast, an enterprise-grade NVMe SSD can exceed 1,500,000 random read IOPS. When designing storage systems, engineers often use RAID configurations or Flash arrays to aggregate IOPS and meet the requirements of high-load applications.