wear leveling

Think of a flight of carpeted stairs. If everyone always steps on the exact same spot of each step, that spot wears bare in a year while the rest of the carpet stays new. A smart caretaker would gently nudge people to step in different places over time, so the whole carpet ages evenly and lasts far longer. Wear leveling does this for flash memory: it spreads writes across all the cells so no single cell is worn out long before the others.

Concretely, each flash erase block can survive only a limited number of erase cycles (perhaps a few thousand to tens of thousands) before it becomes unreliable. If the drive always reused the same physical pages for hot, frequently-rewritten data (say a file system's metadata), those cells would die quickly while most of the chip stayed pristine — and a drive is only as healthy as its weakest blocks. Wear leveling, run by the flash translation layer, deliberately rotates which physical blocks hold which data, even occasionally moving rarely-changed (cold) data to free up lightly-worn blocks for hot data, so that erase counts stay roughly balanced across the whole device.

Why it matters: wear leveling is the main reason modern SSDs last for years of normal use despite flash's limited write endurance. Without it, a drive could fail in months because a few hot spots burned out. But it is a balancing act with a cost: moving cold data around to even out wear is itself extra writing, contributing to write amplification. And it does not make flash immortal — it spreads a finite budget of writes evenly, buying longevity, but a drive hammered with writes will still eventually exhaust that budget.