NVMe Changes the Data Recovery Game

TORONTO — Data recovery companies that started out decades ago salvaging information from spinning disk worried flash and SSDs might put them out of business. Instead, it's created new challenges.

In an interview with EE Times, Mike Cobb, director of engineering for Drivesaver, said pulling data from flash media and SSDs has made each recovery a custom project. He's been with the company nearly 24 years, and started out working on 3.5-inch hard drives as IDE parallel devices were becoming more mainstream since they were cheaper than SCSI. “There wasn't a lot variation until companies went SATA," he said.

Now the hard drive market is dominated by only two players — Seagate and Western Digital — while the SSD segmented has exploded, Cobb said. Players have already come and gone, with companies such as Intel, Samsung and Micron dominating as the market consolidates.

“The first several years for flash were insanely crazy," Cobb said. Controllers, encryption and interfaces all contribute to the data recovery challenge, he added. “Right now, we're in the wild west."

Compared to hard drives, where lessons from one device could be applied to others, pulling data from an SSD is more of an R&D project, said Cobb. “Each SSD in same family could be monumentally different," he said. "It's not so much the media itself. Your NAND is pretty much NAND."

Firmware has always played a role in hard drive recovery, Cobb said, and SSDs are no different. “They have a tremendous amount of overhead," he said.

DeepSpar Data Recovery in Ottawa, Canada, has built its business out of such R&D. Rather than doing the recovery work, the Canadian company partners with major data recovery companies to give them the tools they need as storage media evolves. Sales Director Serge Shirobokov told EE Times the major data recovery firms are just starting to see an uptick in NVMe SSDs. Until last year, most of the SSDs needing recovery have been HDI PCIe drives from MacBook Air computers, he said.

DeepSpar's recently announced tool to help with the recovery of data from failed NVMe SSDs recognizes that a completely different control features are necessary to pull information off failing or damaged devices. One of them is the ability to manually limit the number of PCIe lanes being used by the SSD. This helps restore access to data if particular PCIe lanes have a fault. If there is an electronic problem on lane 3, then it can be avoided by only using lane 1 for all communication. In the same manner, manually limiting the PCIe link speed to the lowest setting can help increase the signal to noise ratio and achieve stable communication with the SSD.

Another control feature is the ability to send various PCIe-level and protocol-level resets upon read timeout or read error, essentially enabling the tool to do a soft reboot of the SSD and continue recovering data in case the SSD becomes unresponsive as a result of attempting to read damaged flash cells. In the same manner, in case resets are insufficient, the tool can do a hard repower of the SSD to continue the recovery process.

Shirobokov said all it takes for a computer not to be able to properly boot is for a few flash cells to fail on the SSD. The DeepSpar tool needs to be used sooner than later. “Our tool let you recover data from an unstable SSD. It will fail completely after some time," he said. “It gives you larger window of opportunity to catch things before it completely fails."

There are similarities and differences between hard drives and SSDs when it comes to device failure and data recovery techniques. DeepSpar's new tool builds on its imager for hard drive recovery, said Shirobokov, but unlike spinning disk, SSDs don't provide indications, such as abnormal noises, that they are close to failure. It just doesn't boot. Flash also fails in a cascading manner.

Both hard drives and SSDs can have bad sectors that lead to data loss, but for the latter, it takes only small percentage to make the drive inaccessible, said Shirobokov, and since the translation tables are stored on flash, the controller can stop working too.

He said it's too early to say what the success rates and recovery times are for DeepSpar's NVMe SSD recovery tool, but the company will likely have some statistics this time next year.

NVMe SSDs are becoming more widespread as they hit prices appealing to consumers, said Cobb, and although the interface will make some aspects easier, each manufacturer will have their own critical components located in different places on the board. “The chances of good recovery are going to be higher on the SSD, but the timelines to get those recoveries are going to be extended," Cobb said. 

Cobb said SSDs have changed what causes data loss. “It's less likely the user will destroy the drive. You can drop your notebook all you want now," he said.

Backups to the cloud have changed the dynamics of data recovery, too. Instead of floppies and hard drives, it's cellphones and USB drives. And although mission critical information is more likely to be replicated to a secondary storage array or the cloud, it seems there's always something that's not backed up, he said. “We don't see the end of data recovery anytime soon," Cobb said. "It's getting more exciting. It's more cerebral."

—Gary Hilson is a general contributing editor with a focus on memory and flash technologies for EE Times.

Related Articles:

• Storage Industry Group Tackles SSD Data Recovery
• Destroying Data Is Hard Work
• NVMe Will Oust SCSI by 2020
• NVMe Spec Gets Major Update
• Does NVMe Have a Place in Industrial Embedded and IoT?