rotational latency
Picture a spinning lazy Susan on a dinner table with dishes around its rim, and you are allowed to grab a dish only as it passes the spot in front of you. Even after you reach for exactly the right spot, you still have to wait for the dish you want to come around. If it just passed, you wait almost a full turn; if it is about to arrive, almost no wait. Rotational latency on a disk is that wait: after the head is on the correct track, the time for the spinning platter to bring the wanted sector around under the head.
Concretely, the platter spins at a fixed speed, so the wait depends only on where the target sector currently is relative to the head. In the worst case the sector has just passed and you wait nearly one full revolution; in the best case it is already arriving and you wait almost nothing. On average you wait half a revolution. You can compute it directly from the spin rate: at 7200 RPM the platter turns once every 60 / 7200 of a second, about 8.33 ms, so the average rotational latency is half that, about 4.17 ms. Faster-spinning drives (10000 or 15000 RPM, in older enterprise disks) cut this delay.
Why it matters: rotational latency is the second of the three components of disk access time, sitting between seek time and transfer time. It is pure waiting — the drive can do nothing but spin until the data arrives — and it sets a hard floor on random access speed that no scheduling can fully erase (scheduling mainly attacks seeks). Some advanced schedulers do try to order requests by rotational position too, but it is delicate. Like seek time, rotational latency simply does not exist on an SSD, since there is nothing spinning.
At 7200 RPM the disk makes one turn every 8.33 ms. If the head lands on the right track just as the target sector has slipped past, the drive waits almost the full 8.33 ms; if it lands just before the sector, almost none. Over many random reads, the average is about 4.17 ms.
Average wait is half a revolution, fixed by the spin rate.
Rotational latency is unavoidable waiting, not slow data transfer. People often lump it into 'seek time', but they are different: seek moves the arm, latency waits for the spin. Total access time adds seek, latency, and transfer.