SSD - larger drives vs. RAID 0. More chips = less reliability?

[bradlygmathews]

Had a thought today while thinking on my new SSD RAID 0 array. I upgraded to improve database performance and keep my data very well backed up.

2 drives increases failure risk - this is the common knowledge. But is that actually true with SSDs?

2 SSDs have twice the number of chips as one. But typically, so does a drive with twice the capacity.

So is my Crucial M4 2x128GB RAID 0 array really more prone to failure than a single Crucial M4 256GB drive?

If we factor in controller chips, cables and the like, then yes, there are more points of failure with multiple drives. But I have to wonder, the the world of spinning platters what percentage of actual "Oh my God my data is fried!" failures are caused by the HDD mechanics vs. the controller chips, cables and the like?

I have had many hard drives die on my over the decades - the first HDD I used was a 5MB beast the size of a full AT case. Somewhere I still have a 10 or 12" platter with a divot in it from a hard drive my dad repaired, and I have seen literally hundreds of head crashed 5.25" platters at hard drive recovery shops. So maybe my experience prejudices me toward assuming most failures are mechanical.

- Brad

 

[beenthere]

In theory SSDs should be more reliable than platter HDs but so far consumer grade SSDs have had a lot of reliability and compatibility issues that they should not have. As far as the physical chips go they should be reliable, but the controller, firmware etc. is where many issues exist.

In my experience of 20+ years building PCs I've had (1) SCSI enterprise quality hard drive fail and many IDE/SATA drives fail. Consumer grade hardware can be real crap rushed out the door for huge profit...

 

[bob100684]

mathmatically, your chances of failure go up exponentially....lets use an example, if a device has a 5% chance of failure, and you use two of them, either one failing taking the system down, much like raid0, you now have not a 10% chance of failure, but a 5%x5% or 25% chance of failure.

 

[sminlal]

Reliability is dependent of how many components and connections there are and the risk factor for each. A larger SSD has more chips, but it doesn't have double the total number of components and connections the way a completely separate drive would. So yes, I can see that there might be slightly higher chance of failure with a higher capacity SSD, but in practice I doubt that an SSD with twice the capacity would have more than perhaps a 10 to 25% additional chance of failure. It certainly wouldn't be twice as much risk overall.

 

[Dereck47]

2 drives increases failure risk - this is the common knowledge. But is that actually true with SSDs?

Yes, that is true with SSDs as well.

But note that with HDDs or SSDs in RAID it is the odds of failure that increase; it does not mean that failure will happen.

If I own 1 2011 Ford Mustang and you own 3 2011 Ford Mustangs; if we both put the same mileage a day on our cars the odds are 3 times greater that 1 of your cars will be in an accident before mine. But you could go 100k on each car accident free and I could get into a fender-bender on day 1. This scenario is a little simplistic with regards to driving habits, but I hope you get the overall point I’m trying to make.

 

[bradlygmathews]

Derech47 and bob100684,

Here is an interesting post on statistics. Not sure which case applies to hard drives. Wikipedia has a formula as well.

http://answers.yahoo.com/question/index?qid=20070817065532AAwXugv
http://en.wikipedia.org/wiki/Standard_RAID_levels#RAID_0_failure_rate

Also let's look at a pic of a SSD PCB:

http://benchmarkreviews.com/index.php?option=com_content&task=view&id=437&Itemid=60

So if we assume an equal failure rate of each component (IC, resistor, diode, crystal) then I guess yes, doubling the number of memory chips on a single pcb only adds 20% more so more failure points vs doubling the failure points with two drives.

However, doubling the capacity of a HDD adds no additional parts (assuming bit density increases vs, additional platters,) you just increase the odds of bad sectors (which have not caused me any known data loss in the last 15 years).

This article is the best I can find on reliability rates for SSDs vs HDDs:

http://www.tomshardware.com/reviews/ssd-reliability-failure-rate,2923-3.html

I have no data on small vs large SSDs.

So averaging the scores gives me these reliability % (using the formula in wikipedia):
Single Drive SSD 98.01%
2 drive RAID 0 SSD 96.05%
Single Drive HDD 96.70%
2 drive RAID 0 HDD 93.51%

Hmm. 2 drive RAID 0 reliability is almost the same as single HDD.

Not the direction I was going with in this post, but very interesting. I need peer review on my data please.

- Brad

 

[Dereck47]

Derech47 and bob100684,

Here is an interesting post on statistics. Not sure which case applies to hard drives. Wikipedia has a formula as well.923-3.html

Thanks!

 

[groberts101]

hardware, power supply, and cabling are all the biggest factors to consider. I use 6 Vertex 2's in R0 on my board(X58/ICH10R) for near 1.5 years now without issue.

No failures.. no boot issues(staying booted 24/7 makes that possible with my I7-920's heavy OC(4.2ghz) including the PCI-E @112mhz) and even without so much as a chkdsk run or smart error. Lost 3 array's while screwing around with Raxco's Fancycache though.. but that doesn't even come close to counting due to heavy deferred write testing at each ocurrance.

However.. I should point out that I always use secure erase and dedicated imaging/recovery protocol's to keep my drives/OS clean, lean, and healthy at all times.

Top line raidcards are best.. but just like a nicely binned CPU or RAM module?.. there are exceptional mobo chipsets out there that will still get the job done with stability to spare.

 

[JohnnyLucky]

Manufacturers do not reveal their statistics about drive issues, problems, and failures so we have no way of knowing about actual reliability.