WAN is an acronym for Wide Area Network. WANs are networks established over a wide area such as over a city, a state, or a country. Generally, WANs include anything beyond a LAN, or Local Area Network.

wear leveling

Wear leveling is a technique that extends the lifespan of an SSD by distributing write operations equally across all blocks of NAND.

This is important because all NAND flash memory produced today have a limited number of erase/write cycles. Once the endurance limit has been reached, the expired blocks would no longer be able to be rewritten with new data.

To preserve the limited number of erase/write cycles, the SSD controller doesn’t actually perform an erase operation on a data block when it’s issued a deletion command. Instead, it simply marks the block as invalid, which simply means that the data can be safely overwritten.

When a new write command is issued, the controller will route that write command to blocks of NAND with the least amount of write/erase cycles. This ensures that all NAND on the entire drive is being degraded at the same rate, allowing the drive to continue to operate until its nearing its end of life.

write amplification

Write amplification is an undesired phenomenon in SSDs when more writes to the NAND occurs than necessary. Due to most NAND having a limited number of write/erase cycles, high write amplification could significantly impact the lifespan of an SSD.

Write amplification could happen in a variety of ways, but the most common way it can occur is the way the SSD handles modifications to a specific block of data. Storage blocks are divided into pages, and pages within the SSD can’t be deleted individually. This means that whenever a few pages need to be altered, all the old data within the block needs to be erased and rewritten to include the modified data, wasting erase/write cycles in the process.

Write amplification can be mitigated with over-provisioning, TRIM, wear-leveling, and garbage collection.

write-back cache

In a write-back cache, writes are confirmed to the host immediately after data is written to the cache. From there, data is then written from the cache to permanent storage typically when the cache is full or when the cache is idle.

As the cache is significantly faster to write to compared to permanent storage, the write-back cache is the fastest type of cache however, the speed of the write-back cache does come at a risk of data loss. In the event of sudden power loss, all data in the cache would be lost, including any data not yet written to permanent storage.

write-through cache

In a write-through cache, writes are confirmed only after the data has been written to both the cache and the permanent storage behind it.

As writing to both the cache and permanent storage is slower than just writing to the cache itself, the write-through cache is significantly slower in comparison to the write-back cache. However, as data is written to both the cache and permanent storage, data is secure even in the event of sudden power loss.