I have a question about the capacitors in the DC S3700 and the DRAM cache. In the event of a power loss while write caching is enabled on the device, do _all_ of the cached/buffered writes make it to the NAND? Or, are the capacitors there to ensure that any active writes at the NAND are written atomically (and prevent corrupting data/metadata/tables/etc)?
I was curious because depending on the answer, I'll turn write caching on or off in our server setup, or purchase the battery-backed option for our RAID card.
I don't see such information in the product data sheet, yet some press releases and reviews imply the first option to be the case.
I found this verbiage on ARK:
Enhanced Power Loss Data Protection prepares the SSD for unexpected system power loss by minimizing data in transition in temporary buffers, and uses on-board power-loss protection capacitance to provide enough energy for the SSD firmware to move data from the transfer buffer and other temporary buffers to the NAND, thus protecting system and user data.
Is there a less-fluffy statement that can affirm that all buffered writes will get to the NAND atomically? (i.e. All commands which have been acknowledged by the SSD with 50h/00h status while write cache is ON will be written on power-loss).
Yes, the Intel® DC 3700 SSD uses capacitors to write the data from the buffers to NAND during a power loss.
"The drive saves all cached data in the process of being written before shutting down, thereby minimizing potential data loss."
For further information you may check the following document:
I've also read those statements in the datasheets and specifications. My concern is with the statement "in the process of being written." This does _not_ state that all buffered data is written.
Can Intel provide a more complete description of what gets flushed upon unexpected power-loss, just as the competitor has listed in the above picture? I'm just concerned because the amount of bulk capacitance on this http://www.storagereview.com/samsung_ssd_sm825_enterprise_ssd_review SSD can power the drive for 55 seconds.
Note that I'm taking your word for it, but if they don't behave like you say they do during power cut testing (by exhibiting errors or shorn writes) we'd be forced to return these drives. Maybe you can run and go check with one of your FTL firmware engineers or FAEs?
Just to confirm, if there are a full queue depth of commands that have been acknowledged with an ATA 50/00h status, and I cut power at random times, I will get correct data back?