SSD Problems caused by power failure:
Generally power failure would cause 3 main problems to SSD when PC runs.
- Mapping table might get lost because there is no enough time to update and store it. Usually SSD manufacturers do not provide users with tools to recover mapping table. Thus SSD will become undetectable to host and only method is to return the drive to manufacturer for repairing. And SSD vendors will reproduce the drives by firmware which will get all data on SSD lost.
- SSD mostly deploy DDR or SDRAM as the cache on board. Once power fails, all data in cache will be lost because both DDR and SDRAM are non-volatile media. And these data have never been transferred into NAND flash yet because R/W speed of DDR is far higher than NAND Flash. However host will consider that data writing has already been finished. In this way power failure causes data loss.
- Power failure results in new nominal bad blocks. ECC of controller checks 256bit data in row and column, and xor each bit(If result is 0, it means even numbers of 1. If result is 1, it means odd number of 1). When power failure occurs, there might be errors above 1 bit in 1 block, which cannot be corrected by ECC because it only deals with 1 bit’s error at 1 time. Then controller will misjudge this block as bad one and will not use it for new data storing. This is why common SSD tools find increased bad block numbers after power failure happens.
Renice deploys super capacitor or tantalum capacitors on specific SSD models, and detect the input power from both capacitor’s and SATA power’s input end. Detection circuit is interfacing with corresponding GPIO so that abnormal signal of power will be reported to SSD Firmware through GPIO when power fails. When 5v power supply breaks off, SSD will report to HOST and stop to receive data not reaching SSD. Renice solution also takes over-current into consideration. Once the input current is detected to excess limitation, current-limiting circuit (Protection circuit to prevent damaging over components if instantaneous input current is too large) will report to SSD controller, as reactions original power supply will be cut off and capacitors will provide the power required to transfer mapping table and data in cache into NAND flash.
Renice’s engineer team developed our own instrument to produce power blackout in different operation process, including reading, writing, shutting down, starting up, rebooting, or even irregular power blackout.
Using this instrument Renice implements power on-off for 3,000 cycles to verify SSD’s power failure protection is able to run well in real operation.