Solid state drive lifespan is the measure of an SSD drive’s usable lifecycle. SSD endurance is based on the number of write/erase cycles a flash block can reasonably accept before producing hard errors or complete failure.
Since most SSDs are made up of NAND flash, the calculations depend on average NAND write cycles. As we look at flash storage memory, we’ll discuss how to make those measurements, and how to prolong SSD lifespans.
SSD Life Expectancy
Throughout the SSD market, SSD vendors rate drive reliability on three factors:
- Age of the drive
- Total terabytes written over time (TBW)
- Drive writes per day (DWPD)
Of the three, only the age of the drive is a precise measurement. Vendors and users can calculate TBW and DWPD, but some of the numbers are estimates.https://o1.qnsr.com/log/p.gif?;n=203;c=204660761;s=10655;x=7936;f=201812281257540;u=j;z=TIMESTAMP;a=20400368;e=i
However, some real-life tests allow us to draw reasonable conclusions. In 2016, Google and University of Toronto reported the results of a joint 4-year study that tracked SSD and HDD reliability in Google’s data-intensive data centers.
The study demonstrated that the age of the SSD was the main factor in drive failures. They also reported that Google replaced SSD drives about a quarter less often than HDDs. However, SSDs reported more errors than the HDDs, with more bad blocks and uncorrectable errors.
Measuring SSD/Flash Endurance
NAND flash SSDs have a limited number of write cycles before the cell fails, expressed as its endurance rating. The cause is physical: every time the drive writes/erases, the flash memory cell’s oxide layer deteriorates. The type of cell impacts the number of write cycles before failure. Notice that when you compare SLC vs. MLC vs. TLC, you'll see key differences:
- SLC: Single-level cell NAND flash supports 50,000 to 100,000 write cycles.
- MLC: The 2-bit data multi-level cell (MLC) flash generally takes up to 3,000 write cycles. eMLC (enterprise MLC) sustains up to10,000 write cycles, and can reach 35,000 cycles on 3D NAND.
- TLC: Triple-level cells (3-bit) NAND flash is low at 300-1000 write cycles, and can achieve 1500-3000 write cycles with 3D NAND.
When you’re thinking about write cycles, also keep write amplification in mind. Writes are not simply single writes in the user or application layers. They are multiple writes for redundancy and crash consistency, where the controller copies data to provide redundancy. Additional sources of write amplification include de-duplication, filesystem writes, metadata and log structures, and garbage collection.
SSD Endurance and Durability: TBW and DWPD
Terabytes Written (TBW)
TBW measures how many cumulative writes a drive can reasonably expect to complete over its lifespan. For example, if the drive is rated for 300 TBW, this is how many writes the drive can take before needing replacement.
Let’s look at the Samsung 850 Pro line, built with 3D V-NAND. Capacities include 128GB, 256GB, 512GB; and 1, 2, and 4TB. Samsung calculates the TBW by drive capacity. It offers a 10-year warranty if the writes are less than TBW maximums.
- 256 GB: 150 TBW
- 512 GB and 1 TB: 300 TBW
- 2 TB: 450 TBW
- 4 TB: 600 TBW
Drive Writes Per Day (DWPD)
DWPD measures how many times a user can rewrite an entire SSD every day throughout its warrantied lifetime. For example, if the DWPD is 1 on a 200GB SSD drive, and the warranty period is five years, then users can rewrite the entire 200GB daily for 5 years before anticipating failure. Users can also calculate TBW from this number: at 200GB a day over 365 days a year, for 5 years, TBW is 365.
If its DWPD is 2 on a 200GB drive with a 5-year warranty, then double the numbers and you come out with 730TB worth of cumulative writes over 5 years.
Making the Calculation
The formula below calculates SSD lifespan. The equation is write cycles x capacity, over SSD factor x data written per year. For example, we'll measure the Samsung 850 PRO, a TLC SSD with 1TB capacity.
Write Cycles = 3000
Capacity = 1TB (1000GB)
SSD factor: real amount of data to actual data written = estimate 5 Amount of data written to the drive per year = estimate 1500-2000GB.
Using these numbers, your equation will be 3000 (write cycles) x 1000 (GB capacity) / 5 (SSD factor) x 1750 (GB written to drive per year) = 342 years.
No one is saying that the SSD is going to be error-free for 342 years, or that the technology will exist anymore. But we can be reasonably certain that assuming these numbers, its oxide layer will last this long.
Mean Time Between Failures (MTBF)
MTBF, or Mean Time Between Failures, is a popular measurement for HDDs but not as meaningful for SDDs. Even with hard drives, MTBF is not a precise measurement. They are statistical evaluations based on small sample sizes. And since no vendor wants to wait five years of nonstop testing before they get the results out, the results are extrapolated from a small period.
This leads to MTBF reports of half a million hours between failures, about 57 years, on consumer hard drives. Enterprise drives with higher volumes of reads and writes will have lower MTBF, but are still measured in many thousands of hours.
With SSDs, the measurement is even less meaningful because they lack the moving parts that causes mechanical failures and HDD’s. SSD reliability using MTBF roughly translates to 2.5 million hours MTBF for an SSD. That may be interesting information, but is less helpful than TBW and DWPD calculations.
SSDs have long lifespans, but as the Google study showed they can have more errors than HDD’s within a shorter time period. New technologies are not only increasing drive lifecycles, but also increasing drive reliability.
- Wear leveling evenly distributes writes on all SSD blocks, so they wear evenly. Wear leveling technologies may be dynamic or static. Dynamic wear leveling maps logical block addresses (LBAs) from the operating system to flash memory. It tracks which block was erased/written last, and writes to a different block. Static wear leveling also maps the LBA to physical blocks, but rotates unchanged blocks to replace written blocks. This is also called global wear leveling.
- ECC (error correction code) software corrects random bit errors that are quite common in NAND flash, and helps to correct bit errors from wear. By correcting both types of errors, ECC lengthens the lifetime of a block.
- Bad block management comprises Skip Block and Reserve Block methods, which identify bad blocks and direct the controller to either skip a bad block and go on to the next good one, or to replace a bad block with a reserved good block.
- TRIM commands are not error checking tools, but do improve performance by immediately wiping deleted pages or blocks. Without TRIM, the SSD controller does not actually wipe deleted data until it is ready to write new data to the same location. Since TRIM wipes upon deletion, write performance improves.
How to Check SSD Health
There are several offerings that can help you keep an eye on the health of your SSDs. It is best to use solutions that are designed for SSDs, since HDD diagnostics and solutions may damage your drive. SSD health check tests include S.M.A.R.T. and Samsung SSD tools like Samsung Magician Software, but there are many on the market.
See our list of the top tools: 9 Best SSD Health Check Tools: Monitoring SSD Health and Performance.
SSD Information from Vendors
It’s not up to users to make their own complicated manual measurements of SSD lifespans. Vendors will calculate SSD lifespans for you, and there are many calculators available online if you want to double-check. And be sure to extend your SSD drives useful life by pairing measurements with wear leveling, ECC, and bad block management.
See our list of the solid state drives for various workloads: The Best and Fastest SSDs