SSD Data Reliability

[Peteman100]

I know SSDs are notable for their mechanical reliability if a laptop is dropped or impacted. But how are they in terms of operating reliability? Are they less prone to "spontaneous" failure than HDDs? Would you say an SSD would be a fair replacement for two drives in RAID 1 in a laptop?

 

[PHiZ]

I had an SSD, and it exploded and killed my family!

Also, http://en.wikipedia.org/wiki/Solid-state_drive#Comparison_with_hard_disk_drives

 

[DonDon]

Most hard drive problems are caused by some kind of mechanical failure in the disk platters, read heads, or the drive mechanisms for the above mentioned areas. While it is possible for the Solid State components on the controller board to fail, that is fairly rare.

SSD's on the other hand have no moving parts. The only mechanical point of failure is the Sata and power connectors, and possibly a bad solder joint. They have a error recovery routine so if an individual memory location should fail, it is automagically detected and replaced with on board spares.

In the long run, they SHOULD prove to be more reliable than mechanical hard drives, but only time and experience with them will tell.

I trust them enough that I am running 4 X 30 gig in Raid 0 on 1 of my systems, and 3 X 60 Gig in Raid 0 on another.

Don

 

[Peteman100]

Thank you both! (sarcasm and all)

 

[Yeorwned]

Drive for drive, spinning drives do fail more often due to long term vibration while SSD drives fail due to a hung deep error recovery attempt in my enterprise test lab. I don't recommend SSDs to anyone in a desktop but in the laptop world, they are quite the blessing for both performance, reliability, and power usage. I would definitely consider displacing the RAID 1 solution in a laptop, especially if it is for normal usage.

I've only seen significant problems with SSDs in laptops where the laptop was used for heavy compiling of c and c++ applications where write cycles to small files can be extremely heavy.

 

[jimhsu]

You must read this, then:

http://www.imation.com/PageFiles/83/SSD-Reliability-Lifetime-White-Paper.pdf

Briefly, if we are talking about read failure (we know a lot about write endurance already), SSDs use an error correcting code (ECC) called BCH. Failure modes for SSDs are very different in that typically single sectors fail in a uniform distribution. Hard disks however most likely have large swaths of failures where failed sectors are adjacent (i.e. mechanical damage). Thus, implementing an longer ECC almost guarantees order of magnitude improvements in unrecoverable error rates for SSDs (see page 14 for numbers).

The Intel X25-M G2 for example guarantees a <10^-15 BER. I assume that doesn't account for controller failure, etc but that's quite good for storage.

Really, despite what Intel says, write endurance is not THAT important for enterprise applications, simply because storage gets obsolete, and even after all cells reach the write limit, the SSD is still readable.

 

[maven]

Contrary to what most people think, under normal usage the disk drive PCB has the highest failure rates, not the HDA.

 

[extide]

Contrary to what most people think, under normal usage the disk drive PCB has the highest failure rates, not the HDA.

I doubt it. Rarely do I have drives fail such that a PCB swap would fix them.

 

[PHiZ]

Contrary to what most people think, under normal usage the disk drive PCB has the highest failure rates, not the HDA.

I doubt it. Rarely do I have drives fail such that a PCB swap would fix them.

Well, I suppose we could consult a data-recovery shop for hard numbers, but if they primarily do PCB replacements, they might not want to admit it. It's probably easier to charge $1200+ for data recovery when people assume you are doing some intensive platter voodoo.

 

[Peteman100]

This is all great info. Thanks again!