What Is IOPS
IOPS Defined: Counting Storage Operations per Second
IOPS, short for input/output operations per second, is the number of read or write requests a storage device completes in one second. It shows how many separate pieces of work a disk, SSD or storage array can handle, regardless of how much data each request contains.
Each operation reads or writes one block of data, and the block size changes the result. A drive completing 10,000 operations per second moves about 40 MB per second with 4 KB blocks. With 64 KB blocks, the same count moves about 640 MB per second. This is why an IOPS figure should always be read together with its block size.
Random vs Sequential I/O and Why Busy Sites Depend on It
IOPS matters most for random I/O, where each request points to a different part of the disk. Sequential I/O reads or writes data in one continuous run, such as streaming a large video file. Hard disks handle sequential work well, but random requests force the head to move, which typically limits an HDD to a few hundred random operations per second. SSDs deliver many thousands. The difference is covered in SSD vs HDD.
Several common workloads generate random I/O:
- Databases, which jump between index entries and table rows for every query.
- Mail servers, which store many messages as separate small files.
- Busy content sites, which write session data, cache entries and logs continuously.
How IOPS Limits Shared Hosting Performance
On a shared server, many accounts use the same disk, and their random requests join one queue. A tenant that uses more than its share of IOPS, sometimes called a noisy neighbour, can slow every other site on that disk, even if those sites receive little traffic. Hosts monitor storage activity for this reason and may limit how much I/O each account can use.
IOPS and latency are closely linked. If each request takes longer to complete, fewer requests finish per second, unless the drive can process several at once. The relationship between the two is explained in IOPS vs throughput vs latency.
Frequently asked questions
What is a good IOPS number?
There is no single good number. A figure only makes sense next to its block size, its read and write mix and the queue depth used during the test. Vendors often quote peak results from ideal conditions. For hosting, the useful question is whether the storage keeps up with your busiest database and file activity.
Does higher IOPS always mean a faster website?
Not always. Much of a website's time goes to running code, waiting for the network and serving cached content. IOPS matters most when the disk is the bottleneck, such as a database under heavy write load. Caching can reduce the random reads a disk must serve, which is explained in SSD caching.