6 common misconceptions about SSDs

Tuesday, 30 July 2013 by Ontrack France

6 idées reçues fréquentes sur les disques SSD

SSDs will never fail and can be easily repaired? It's wrong. We hear all kinds of misconceptions about SSDs ... Here is a brief selection.

SSDs have no moving parts and therefore very rarely fail.

Since SSDs have no moving parts, they are in some ways much more reliable than traditional hard drives. However, they are mainly installed in mobile devices, which are subjected to much harsher conditions than the uses of traditional hard drives. According to Ontrack's observations, water is the most common cause of SSD failure. It causes short circuits and damages electronic components. By comparison, less than 1% of traditional hard drive failures are due to water.

You don't need to take a backup if you sync your SSD to the cloud

There is a rule when it comes to IT: if you don't have a backup on 2 different media, you don't have a valid backup. This rule does not change anything just because the storage medium you are using is more reliable than another. Ontrack often helps customers recover their data after versions of a file saved on the computer and in the cloud both become inaccessible, corrupted, deleted, or overwritten. Very often in fact, the local version gets corrupted and then syncs to the cloud. While the majority of cloud providers save previous versions of files, this solution does not always work. We always recommend doing a second backup.

SSDs can be easily repaired if something goes wrong

This is wrong, SSDs are difficult to separate from other electronic components and are therefore very difficult to repair. For example, SSDs on most tablets and smartphones are nested on the main circuit board.

If the computer has not been turned off properly or the battery is completely depleted, it will not damage the data.

In order for an SSD not to lose data when the power is removed while it is copying data, the SSD must shut down in a safe state within milliseconds. It is taking a risk to hope that everything goes well each time and that the designers of the controller have done their calculations correctly. The experience of Ontrack engineers shows that sometimes things go wrong. The observed results are an alteration of the data to the point that the user can no longer have access to his files.

SSDs will still work because they don't wear out

This is not true, as the lifespan and capacity of all media decrease over time. A design phenomenon called "write amplification" is that the capacity and speed of SSDs are gradually reduced. This phenomenon is caused by the way the SSD erases data from the memory.

The useful life of an SSD can be as little as five years if it is used often. It is advisable to modernize your devices every three years in order not to have this problem. Two types of memory are currently used in SSDs: SLC (single-level cell) and MLC (multi-level cell). SLC memory lasts longer but costs more to produce. All inexpensive drives and most USB Flash memory drives use MLC memory. Flash memory can only be programmed and erased a limited number of times. This is called the maximum number of write / erase cycles that flash memory can endure during its lifetime. SLC memory, designed for superior performance and greater endurance, can generally accept between 50,000 and 100,000 cycles. MLC Flash memory, on the other hand, is designed for lower cost applications and significantly fewer cycles, typically between 3,000 and 5,000 (2011 figures).

There are no major drawbacks to using encryption

This is wrong, encryption greatly reduces the possibility of data recovery in the event that a hard drive is reformatted by mistake. Using full disk encryption complicates this problem. Many data loss situations caused by logical failures cannot be recovered. If the failure damages the encrypted disk and prevents mounting of the disk, the data cannot be decrypted. In addition, if the disk is badly damaged, it is also not possible to attempt partial or targeted recovery as the location of important data cannot be identified. It's an all-or-nothing case of recovery.