Reliability of Solid-State-Disks (SSD)

As we know, the NAND type of memory only allows a finite number of writes. This means that the most pressing issue regarding the reliability of SSD drives (and NAND technology in general) is the life expectancy and durability.

The reason for the finite number of writes is related to the process of storing information. The NAND stores the data in flash regions called ‘floating gates’ that are capable of storing the information by holding the electron charge. Unfortunately, every time we write data into these floating gates, the process of writing tends to diminish slightly the conductive properties of the region, which in turn lowers the ability to hold a charge.

The SSD manufacturers are trying to correct the issue by cycling through floating gate regions (memory blocks) and wear them out evenly. They are also finding the new ways of writing the data, to decrease the degradation of memory blocks. However, after a certain number of writes, the SSD drive will eventually experience partial or completely failure.

But let me illustrate this problem a bit… Recently, I bought a new SSD drive for my desktop computer. Looking at the technical specifications outlined on the packaging box, it listed MTBF. The value of MTBF directly speaks to SSD’s reliability and the length of time until complete failure.

MTBF or The Mean Time Between Failures refers to the failure rate of the drive over its expected lifetime. MTBF tests are done by running simultaneous tests on multiple SSD devices at the same time (typically +1,000 drives). Over time, the average number of failures is averaged and thus MTBF can be calculated.

In my case, MTBF was listed at 2,000,000 hours, which means that I can expect to see my SSD drive to die in approx. 228 years. That may look like a rather large number, but this is not the time until I will see the failure of the first memory block. It’s an average duration until my entire drive, and all of its memory blocks will fail. So, if my drive’s MTBF lists at 228 years, does it means that 1% of the drive’s capacity will fail within 2.28 years? Not really! The rate of memory block degradation will grow over time, meaning the memory block capacity for storing information will be exhausted gradually.

Currently, we see the trend of using CTF (Charge Trap Flash) manufacturing process in NAND memories. CTF is using “a silicon nitride film to store electrons rather than the doped polycrystalline silicon typical of a floating gate structure.” (Charge trap flash, 2016). This process offers many benefits over floating gate technology. The main advantage is related to reduced geometry which not only reduces the total size of the memory chip but also its cost. Another benefit of the CTF technology is tied to ability of packing multiple bits into a single memory cell, which creates better density and allows for more data to be stored on the single chip. These are the main reasons for improved reliability.  CTF technology not only has a potential to succeed over floating gate structure, but I believe the trend is already visible, as per Wikipedia, charge trapping flash accounted for 30% of 2008’s $2.5 billion, and that is just in the NOR flash market. As far as future, the new V-NAND chips will use the same Charge Trap Flash (CTF) structure with memory cells  stacked to form a 3D model vertically forming a 48-stories tall stack.
“In total, each chip contains over 85.3 billion cells. They each can store 3 bits of data, resulting 256 billion bits of data, in other words, 256Gb on a chip no larger than the tip of a finger.” (Size, F. , 2015).

 

To conclude, the life expectancy is the biggest reliability problem of SSD drives, because it guarantees a failure. More writing cycles directly equal to the speed of failure. If you wonder what is the best measure of life expectancy of an SSD drive, I would suggest going by the manufacturer’s warranty. In a case of my SSD drive, it was listed at 10 years, meaning I shouldn’t experience any issues within the first 10 years of usage.

 

References

IOPS (2016) in Wikipedia. Available at: https://en.wikipedia.org/wiki/IOPS (Accessed: 21 August 2016).

Slack, E. (2012) Available at: http://www.storage-switzerland.com/Articles/Entries/2012/3/6_Why_Flash_Wears_Out_and_How_to_Make_it_Last_Longer.html (Accessed: 21 August 2016).

Gasior, G. (1999) The SSD endurance experiment: Data retention after 600TB. Available at: http://techreport.com/review/26058/the-ssd-endurance-experiment-data-retention-after-600tb (Accessed: 21 August 2016).

Charge trap flash (2016) in Wikipedia. Available at: https://en.wikipedia.org/wiki/Charge_trap_flash (Accessed: 22 August 2016).

Size, F. (2015) Samsung electronics begins mass producing industry First 256-Gigabit, 3D V-NAND flash memory. Available at: http://www.businesswire.com/news/home/20150810006386/en/Samsung-Electronics-Begins-Mass-Producing-Industry-256-Gigabit (Accessed: 22 August 2016).

Facebook Comments