OVER PROVISIONING

Hard disks and SSD drives both provide users with more storage than it presented to the user. This allows vendors to guarantee some given amount of storage. The spare storage is held in reserve in case of media errors.

Hard disks use spare sectors on a given track so that if a bad spot is noticed it can be swapped out. The move to 4096 byte sectors allowed hard disk vendors to use better block sizes for more robust error recovery. In practice, the more modern 4096 class disks have been far less likely to present defects which affect users.

aa22aa138315776a4d94b2142c21d21a

Logical Block Address (LBA) is a simplified API where all that is needed is address of the the sector to be read and how many sectors in total are desired. LBA has allowed hard disk vendors to change the design of disks to make them more robust. Spare blocks can be arbitrarily used.

SSD products started off as a cache for hard disks. Servers started using SSD products for databases and high traffic web servers.

Over time the capacity of the SSD made them suitable for an operating system. Laptop users began buying vast numbers of them and replacing the OEM hard disks. Eventually the capacity of SSD products reached the hard disk realm.

SSD designs also use spare blocks. The spare blocks are present so that when a sector is damaged it can be reallocated elsewhere. With the original NAND now being replaced by MLC, TLC and now QLC the capacity has exploded making the SSD more useful corporate needs.

SSD has a wear problem that was discovered in corporate servers. Analysis suggested that there was some unforeseen factors. Vendors redesigned their products with spare blocks and wear levelling to improve on the service life. The controller logic became a crucial feature as NAND chips moved to becoming a commodity.

The real revolution with TLC and QLC has been the development of the M.2 standard where a SSD is mounted on the motherboard. Capacity ranges from about 32GB though 2TB. Realistically for Windows, a 128GB SSD, or larger, is really a necessity due to the overhead of updates etc.

The introduction of stacked NAND memory has been heralded as a move towards overcoming the limitations of semiconductor fabrication. It started with dual layers and soon after 4 layers in a bid to increase capacity.