The Week In Storage: Self Serving Survey Says SSDs Are Screwy, NVMe Over Fabrics Released

[Guest]

We climb atop the storage soapbox to rant and rave about all things storage.

The Week In Storage: Self Serving Survey Says SSDs Are Screwy, NVMe Over Fabrics Released : Read more

 

[sh4dow83]

That survey does seem questionable to say the least. Personally, I had about six HDDs fail on me over the years but no SSD so far. I own 3.
And so I plan on migrating my HDD RAID1 to SSD once the prices drop a bit further still.

 

[sh4dow83]

By which I'm by the way not claiming that MY experiences are representative of anything.
By the way... why didn't you guys research the average failure rates of HDDs? Because I bothered doing that before relying so heavily on SSD and found that they indeed fail much more often.

 

[John Lauro]

I have lost many hard drives and ssd drives. Not sure which I would say is more reliable, but what I have found is HDD can tend to be brought back to life a little and read most if not all the data. However, when a SSD decides to fail (and I am talking failure way prior to it's estimated life expectancy), it dies immediate, no way of setting up a machine to keep retrying and eventually getting the data off. Plus the reuse different bits and so it's also less likely a recovery service will be able to pull the data off the individual chips and put it back together.

I think manufacturers should be learning how to have drives fail in a more readable state, but I haven't seen that yet with SSD. I currently have about 120 SSD and way more than that HDDs, and haven't lost any data from a SSD failure. However, all the SSDs are in RAID 1 or RAID 5 or 6, etc...

Based on my experience, HDD give more warning signs they are degrading... SSD just die for no good reason. As long as you have automatic redundancy in place, the failure rates are not that significant even if one was 10x as likely to fail as the other. However, if you only have a single drive in a computer it can be critical...

 

[PaulAlcorn]

By the way... why didn't you guys research the average failure rates of HDDs?

Unfortunately, much of the data that is available is anecdotal, and/or drawn from inherently flawed test environments, which are typically not actual test environments. An accurate measurement of failures would require a large sample base (hundreds) and a solid test methodology and environment. Of course, HDD vendors could just share the failure rate data as a few of the SSD vendors have (Intel, OCZ), but we have not seen that as of yet. Better yet, Google (or any of the hyperscalers) could share their data. That would be the best answer.

 

[CaedenV]

It would be interesting to see how many people have lost data on a SSD 'lately'

I mean, of early adopters, who didn't loose data on an SSD a few years ago? They were crap, and we all knew it, and we all prepared for it. The point was that while it was not reliable, the convenience of speed far outweighed that fact of life.
But, of all the SSDs I have bought in the last 2 years (~12 of them... I have a problem lol), I have only had 1 fail. All of these are in use every day in either my own machines, or machine's owned by close friends. They run the gamut on quality, and the only one that failed is a 2 year old Kingston (which I think is under warranty).
But in that same 2 year period I have also bought 10 3TB HDDs. 3 of which showed up dead, another of which died after 2 months, and another died after a year. Granted, these were all bottom-barrel HDDs (next version of my home server will focus more on quality drives... I just needed to get something up and running), but still... of all of the crap OCZ and Mushkin drives I have bought over the years, not a single one of those have died in the last 3 years.

I know it is anecdotal, but SSDs seem MUCH more reliable than HDDs in this day and age. But, if asked about data loss... it is true, the SSD that died took some important data with it to it's grave, but as all the HDDs have been in RAID I have not lost anything on them. I suppose when you know you can't trust something, you take appropriate precautions.

 

[stevenrix]

From a personal experience i've been using hard-drives for the last 30 years: all my drives failed at some point in time. The fastest failure rates was with Maxtor and Seagate. Your mileage may vary, but some of my drives lasted barely 1 month and other lasted more than 10 years. I switched to SSDs 5 years ago and so far they haven't failed on me yet but that will come one of those days.
It's very important to do backups or have a redundant solution, the price to recover 3 hard-drives of 500 gigs for a company is around $7600. That's a lot of money.

 

[PaulAlcorn]

From a personal experience i've been using hard-drives for the last 30 years: all my drives failed at some point in time. The fastest failure rates was with Maxtor and Seagate. Your mileage may vary, but some of my drives lasted barely 1 month and other lasted more than 10 years. I switched to SSDs 5 years ago and so far they haven't failed on me yet but that will come one of those days.
It's very important to do backups or have a redundant solution, the price to recover 3 hard-drives of 500 gigs for a company is around $7600. That's a lot of money.

Excellent insight, we definitely always need some form of redundancy.
I tried to upvote your comment, but accidentally downvoted it, and it will not let me change it. Gah! I upvote this comment

 

[PaulAlcorn]

I keep 3HD with one Crucial MX200 500GB for programs then another to store all my files,
then i back up everything on my trusty WD Black 2TB HD,
then i Re Back everything up thru online backup,
then take my very most important documents putting them on a thumb drive encrypted then placed in a safe in my underground bunker.

People should never underestimate the importance of the secret bunker in a well-rounded data protection plan.

Also, the cloud makes backup so easy now, it's hard to justify not keeping an offsite backup. I used to take a few Bitlocker'd drives over to a friends house every few months for off-site, but now just uploaded everything to AWS.

$59.99 yearly for unlimited cloud backup (I scored it for $5.99 on a Black Friday special) - Bezos, that was a bad idea, I have 13TB uploaded and counting

 

[ohim]

I had only 1 HDD failure since 2005 (only Seagate HDDs, no WD). And 1 SSD failure, Corsair brand.

But each time i have valuable data i never store it on only 1 drive, i always do multiple back`ups on multiple drives.

The article is misleading on this matter fueling the fear of loosing data. When you have only 1 copy of your data.. expect that there is a possibility that you`ll loose it all!

 

[Darkk]

From a personal experience i've been using hard-drives for the last 30 years: all my drives failed at some point in time. The fastest failure rates was with Maxtor and Seagate. Your mileage may vary, but some of my drives lasted barely 1 month and other lasted more than 10 years. I switched to SSDs 5 years ago and so far they haven't failed on me yet but that will come one of those days.
It's very important to do backups or have a redundant solution, the price to recover 3 hard-drives of 500 gigs for a company is around $7600. That's a lot of money.

Excellent insight, we definitely always need some form of redundancy.
I tried to upvote your comment, but accidentally downvoted it, and it will not let me change it. Gah! I upvote this comment

In enterprise environments you should be running offline backups AND redundancy. Redundancy is great for quick recovery but if you get hit by some crypto ransom malware it's not going to do you bit of good if you don't have offline backups. Don't rely on volume shadows / snapshots as it might get deleted by the malware.

In addition you should have multiple timed offline backups. Finally test your backups!

 

[Drummerdude1099]

I don't mean to distract from the point of the article, but perhaps you meant "Cue the end-of-the-SSD-world headlines" in the 5th paragraph. Good article nonetheless.