Have MLC flash reliability concerns truly been addressed? How diligent do users need to be when it comes to tracking writes to MLC drives?
First, let's clarify what the concerns about flash reliability are. Essentially, all flash types (SLC, eMLC, MLC and TLC) have the same weakness; a finite number of times that they can be written to. The more bits per cell, as is the case with MLC and TLC, the sooner this maximum number of writes is reached.
MLC is becoming very common in enterprise storage offerings. And vendors are claiming they can mitigate these wear-out concerns, while still hitting the right price points to make flash affordable.
As to how much of a concern MLC reliability is, it depends on the use case and how well that MLC drive is protected. The higher the write activity is, the greater the potential for that concern.
For example, an MLC-based cache is more susceptible to failure than an all-flash array. This is because the cache, because it is smaller in size, will constantly be refreshed with new data. The all-flash array will store data, so some of it will be less active.
Another factor is how well protected that MLC flash is. If it fails, is there redundancy in the storage system, such as RAID or mirroring, so that the data can be recreated? If so, that can make flash reliability less of a concern.
This was first published in January 2014