IOPS (Input/Output Operations Per Second) counts how many individual read or write requests a storage device can handle each second; throughput measures how much data moves per second, usually in MB/s or GB/s. A drive can post excellent throughput and mediocre IOPS, or the reverse — the two numbers describe different bottlenecks, and different jobs care about different ones.
What IOPS Actually Measures
Every time an application asks storage to read or write a chunk of data, that is one I/O operation. IOPS is simply how many of those requests a device can complete per second. It matters most when data is scattered in small, random pieces — a database looking up thousands of unrelated rows, a virtual machine's boot disk, or a NAS serving many small files to several users at once. Each request carries overhead (finding the data, opening a connection, confirming the write) that has nothing to do with file size, so IOPS is really a measure of how fast a device handles lots of small errands.
What Throughput Actually Measures
Throughput measures raw data volume moved per second, and it dominates when I/O is sequential — one large, continuous read or write. Copying a folder of 4K video files, backing up a full drive image, or streaming a single large export are throughput-bound tasks: the device spends its time moving bytes, not juggling separate requests. A single big file read can saturate a link's throughput ceiling while barely touching its IOPS ceiling.
Why a Drive Can Be Strong in One and Weak in the Other
Mechanical hard drives illustrate the split clearly. A 7,200 RPM HDD like the WD Red Plus 8TB can sustain solid sequential throughput — well over 150 MB/s reading one big file — because the read head just tracks a spinning platter. But its IOPS is limited to roughly 100–200 random operations per second, because every random request means physically moving the head and waiting for the platter to rotate into position. An SSD flips that relationship: no moving parts means random IOPS in the tens of thousands even on modest drives, while throughput is capped by the interface (SATA, NVMe) rather than mechanics.
- Spinning HDDs: good sequential throughput, weak random IOPS — seek time is the bottleneck
- SATA SSDs: throughput capped around 550–600 MB/s by the SATA III interface, but IOPS jump by orders of magnitude versus HDDs
- NVMe SSDs: both numbers climb — multiple GB/s throughput and IOPS in the hundreds of thousands, because the PCIe interface removes the SATA ceiling
Where This Shows Up on a NAS
Network-attached storage adds a second ceiling on top of the drives: the network link itself. A multi-bay NAS like the Synology DS423 can combine several drives into an array that comfortably exceeds what a single drive delivers, but a 1 Gigabit Ethernet connection tops out around 125 MB/s of throughput no matter how fast the drives inside are. That ceiling is why higher-end NAS units add 2.5 Gigabit or faster network ports — the drives were never the bottleneck once you moved past a single spinning disk. IOPS on a NAS matters most when several people or applications hit it with small random requests at once, such as multiple editors browsing thumbnails in a shared media library; throughput matters most when one person is copying a large project folder or a big export file. For more on matching a NAS to that kind of workload, see our guide to choosing a NAS for photo and video work.
Which Number Should You Actually Care About
Match the metric to the job, not the marketing spec sheet:
- Editing large video files, transferring big backups, streaming media: throughput is the number that predicts real-world speed
- Running a database, hosting virtual machines, serving many small files to many users: IOPS predicts responsiveness far better than a headline throughput figure
- General home or small-office file storage: both matter a little, but the network link and drive type (HDD vs SSD) usually decide the outcome before either spec sheet number does
A useful gut check: if the workload description contains words like "stream," "copy," or "export," think throughput. If it contains words like "many users," "database," or "random access," think IOPS.
Frequently Asked Questions
Does higher IOPS always mean a faster drive?
Only for workloads dominated by small random requests. A drive with huge IOPS but a modest interface can still lose a large sequential copy to a drive with lower IOPS but a wider throughput ceiling — the two numbers answer different questions.
Why does my NAS feel slow even with fast drives inside?
The network connection is very often the real ceiling. A 1 Gigabit Ethernet port limits throughput to roughly 125 MB/s regardless of how many fast drives sit behind it, which is why multi-bay NAS units built for heavier workloads increasingly ship with 2.5 Gigabit or faster ports.
Do SSDs make IOPS irrelevant?
No — they just raise the ceiling dramatically. IOPS differences between SSD types (SATA vs NVMe, consumer vs enterprise-grade) still matter for database and virtualization workloads, just at much higher numbers than HDDs ever reached.
Is RAID a way to increase IOPS or throughput?
Both, depending on the RAID level — striping data across multiple drives can multiply throughput and, to a lesser extent, IOPS, though the gain depends heavily on the specific array configuration. See our plain-English breakdown in What Does RAID Stand For?
Share this article: Twitter