0

I am working on a embedded system where there can be unexpected power losses. So far it has been using a Corsair F80 SSD with Windows Embedded Standard 7 (NTFS), and after a unexpected power loss there were some issues that required the OS to be reinstalled. This is not ideal.

This may not be the right spot for this question, but I can imagine this is an issue that needs to be handled in data centers etc as well.

My question is thus; what SSD would you recommend using, that has sufficient protection against power loss to ensure that a clean boot can be expected when power is regained? There is no need for high write/read speeds.

I have been reading about Intel 710 and 320, as well as Seagate Enterprise SSDs, Kingston SSDNow E100 and Samsung 840 Pro, but the details on the power loss protection is in most cases quite lacking.

Walkingsteak
  • 3
  • 2

3 Answers3

2

Try to protect your environment if at all possible. UPS is the first line of defense. But if this is an appliance, kiosk, or other system where you may not be able to control conditions, you should change your SSD approach.

You'll want to use a purpose-built SSD with a super-capacitor (supercap) in this situation. Intel may have a few models with supercaps, but I typically defer to the offerings from sTec. Look into something like the Mach16 if extreme performance isn't required.

The ideal drop-in replacement for conventional HDDs in embedded systems, data centers, storage arrays and computer-related environments.

ewwhite
  • 194,921
  • 91
  • 434
  • 799
  • Thank you. I will check out what they have to offer. Somebody also tipped me about Enhanced Write Filter, so I will see if this can be of any help. – Walkingsteak May 07 '13 at 13:24
1

I would give you a suggestion from my experience, when you use SSDs in the production server, try not to put OS on them, the best option would be to use enterprise grade SD card, also try to tweak a lot of housekeeping stuff which is writing heavily to the disk symlink it from / to /ram or shmfs. That almost completely will resolve issue of SD card degrading its cells for writes.

As per SSD as OS holder on enterprise systems/servers prevention of corruption would be adding some kind of ring buffer (nvram) card and use as the inter point to buffer the writes and if power is lost replay them on reboot.

On direct question of yours if you have a solid budget on this for your problem you can look into these drives -> Intel 320 Series.

http://www.intel.com/content/www/us/en/solid-state-drives/ssd-320-series-power-loss-data-protection-brief.html

Hope this helps.

Danila Ladner
  • 5,241
  • 21
  • 30
1

The power loss protection measures in SSDs are limited to their own cache. When SSDs use a volatile SDRAM-based write cache as a measure of performance optimization, appropriately sized supercaps would provide power for long enough to get the contents of this cache flushed into the non-volatile NAND cells of the disks.

This however would not necessarily protect your OS from being unable to start after a power outage. While NTFS supports journaling for metadata to ensure that the filesystem remains in a consistent state even after an outage event, the data itself is not protected by the journal and is flushed to disk after a certain delay (mainly for performance reasons so writes would benefit from the cache).

If your disk is not badly designed and would return on synchronous writes even when the data is only written to volatile memory, you should be able to alleviate the problem by disabling the write cache for the disk in Windows device manager.

the-wabbit
  • 40,319
  • 13
  • 105
  • 169
  • Thank you for the response. I understand how this is tricky. Since performance really isn't an issue, disabling the write cache may be a good starting point. From what I understand, most solutions to the power loss protection is just ensuring that the disk accepts no more writes, and has enough power stored in capacitors to finish off writing what's already in cache. – Walkingsteak May 07 '13 at 12:56