Having given some thought to the endurance test, we’re planning to run it in stages and vary parameters each time. The basic idea is to write a huge file to the disk, completely filling it, then read the file back again to check for errors, recording SMART health stats each iteration. This process will then be repeated until either the SSD fails, or it is time for the next stage. We anticipate 4 stages currently, two with the disk empty and two with it containing a file occupying 50% of its capacity. In each pair of stages, one will run with the partition occupying the whole disk, one with 9% set aside for over-provisioning. Each stage will run for at least 150TBW, the claimed life of the disk.
At the end of each stage, we will check on bad blocks and sector reallocations, wear levelling and write amplification amongst others. As the disk ages, we will also report on performance, to try to identify whether this suffers as the flash degrades.