Backblaze data shows SSD failure patterns on trend with HDDs

[Jimmy2x]

A hot potato: For a while now, solid state drives have a become the preferred option over hard disk drives due to a lack of moving components, their inherent magnetic resistance, ability to withstand vibration, and overall durability. However, a recent post from cloud storage provider Backblaze reveals SSD failures trending every bit as high as their HDD counterparts.

Often favored for their increased speed, decreased footprint, and enhanced durability, SSDs have become an industry standard storage alternative for everyone from weekend gamers to data centers. A recent blog post by Backblaze's Andy Klein, however, tells a much different tale regarding their purported reliability and longevity.

Backblaze is a cloud storage and backup provider with four data centers across the US and Europe, and they rely on both SSDs and HDDs for server boot drives, log storage, system diagnostics, and various system reporting tasks. Both types of drives serve the same function in Backblaze's environments and support continuous writing, reading, and file deletion in addition to server startup operations.

In the company's most recent annualized failure rates (AFR) data, they compared statistics for SSDs and HDDs performing the same workloads in the same cloud storage environments. The data tells us that SSDs do in fact experience failure at a rate far lower than the HDDs, however, Klein highlights the fact that many of the HDDs referenced were several years older than their SSD equivalents.

Backblaze's AFR data, which displays failure rates for HDDs beginning in 2014 and SSDs in 2018, reveals a remarkably similar pattern for HDD failure rates (from 2014 through 2017) and SSD failure rates (2018 through 2021). This initial trend paves the way for an interesting question: will the SSD failures continue to trend in a similar fashion? If so, we would expect to see a spike in Backblaze's SSD failure rates in the later quarters of 2021 and continuing for the next several years.

Backblaze regularly publishes findings regarding data center drive failures. While the evidence is far from conclusive, it does raise the question if SSDs are truly more reliable and a better investment than HDDs for specific workloads and scenarios. Until more information becomes available which we'll certainly be reporting back to you, the debate will rage on and the decision will continue to be driven by disk requirements, failure rates, and procurement costs.

Image credit: Thomas Ulrich

Permalink to story.

https://www.techspot.com/news/91543-backblaze-data-shows-ssd-failure-patterns-trend-hdds.html

 

[Superconductor]

Bad with trend line interpretation much?

 

[Squid Surprise]

I don’t see how this shows SSDs failing as much as HDDs... even accounting for age, the % seems to be in their favour...

 

[Dimitriid]

Bad with trend line interpretation much?

Well it is backblaze: even a tiny bit of research will let you find they do this kind of thing all the time: Take consumer hardware and then put it to worst-case-scenario data center loads: of course it's going to fail quicker and more often and it serves the company well since their business is predicated on selling you backup solutions.

 

[seeprime]

I don’t see how this shows SSDs failing as much as HDDs... even accounting for age, the % seems to be in their favour...

At the moment that is indeed the case. When more SSD's reach their write endurance limits, and start failing in quantity, say in a few more years, we'll have better usable data, presuming they list the actual drives they used and the drive model is still available and is unchanged. The biggest drawback to taking advantage of the Backblaze data has always been that the currently availabel drives are not always the same ones they report on.

 

[Alfatawi Mendel]

I'm more concerned with the reportedly sudden SSD failures. Is it true that they will bite the dust without so much as a 'bye'? Saying this with a 4 years old Samsung and 2 yrs old Kingston SSD drives in my system. (slight panic setting in)....

 

[Bullwinkle M]

I'm more concerned with the reportedly sudden SSD failures. Is it true that they will bite the dust without so much as a 'bye'? Saying this with a 4 years old Samsung and 2 yrs old Kingston SSD drives in my system. (slight panic setting in)....

It is of no concern at all

SSD's are inexpensive

I have several spare SSD's available for any sudden problem that comes along

If you do not have a spare SSD yet, then your data must not be very important to you

 

[Vulcanproject]

In theory SSDs will continue to improve in reliability as they develop and grow in size. Enterprise flash drives are still relatively young technologies compared to mechanical equivalents. The explosion only really happened in the late 2000s, and there was a lot of garbage flying about at that time with everyone jostling for market share and quality was compromised. Despite all this the inherent reliability of these drives is clear on the graph.

 

[hahahanoobs]

I'm more concerned with the reportedly sudden SSD failures. Is it true that they will bite the dust without so much as a 'bye'? Saying this with a 4 years old Samsung and 2 yrs old Kingston SSD drives in my system. (slight panic setting in)....

Don't expect a "bye" when it comes to your data. If you care about it then you'll make backups Just like we did with HDDs. Just keep doing what you've been doing until told otherwise.

 

[brucek]

It is of no concern at all

SSD's are inexpensive

I have several spare SSD's available for any sudden problem that comes along

If you do not have a spare SSD yet, then your data must not be very important to you

If your data is important to you, what you need is an effective multi-tier backup system. An empty spare SSD is not part of that system. Amazon, Newegg, etc. have all the spare SSDs I'll ever need but that doesn't mean I'm unconcerned about drive reliability.

My personal life is not a 24x7 managed data center with techs standing by to immediately replace any hardware that fails, and while my important data is on fault-tolerant storage backstopped by a multi-tier backup system its not like every boot drive, every game, every little app is also covered that way. I'm technically proficient but rebuilding and reinstalling a desktop or laptop is at best an inconvenience but also potentially a problem disruption if it happens at the wrong time or place. And while I agree SSDs have gotten affordable, those same prices would seem less so if you mentally felt had to buy two or three of each to have the reliability you want.

Bottom line: drive reliability is a big concern to your average consumers, most of whom will regard drive failure as an unwanted expense and a frustrating technical support exercise.

The good news is to me those graphs show that SSDs are in fact pretty reliable, especially if you assume the Backblaze units are probably seeing heavier workloads than your typical home use unit.

 

[whateversa]

I have never had a SSD fail - I use them for their appropriate use case (I don't exceed TBW expectations).

My 9 year old 90GB Corsair SSD still working
My approx 7 year old Transcend 250GB still working
My 3 year old 550GB samsung Evo still working
My 1.5 year old 1TB HIKvision drive also still working

 

[Fastturtle]

If we see a similar failure trend in the back blaze data, then we need to begin looking for the cause, which I suspect will turn out to be the sata protocol itself though it's also possible that it could be related to design failure of the electronics that was not accounted for and easily fixed by a redesign

 

[kiwigraeme]

I have never had a SSD fail - I use them for their appropriate use case (I don't exceed TBW expectations).

My 9 year old 90GB Corsair SSD still working
My approx 7 year old Transcend 250GB still working
My 3 year old 550GB samsung Evo still working
My 1.5 year old 1TB HIKvision drive also still working

As long as I don't buy the ones with the cheap memory &/or controllers - I have no problem trashing them - doing encoding - whatever . Same as my phone - sits in my pocket , with my keys , nuts , bolts, tools whatever other stuff I throw in there . I do attach a glass screen protector and rubber/plastic/gel casing of some type . They are not my best crockery to only be brought out for the Queen .
I'm not going to swim with my phone in the ocean if it was waterproof , or toss in sand etc .
SSD have been tested into the ground - and hold up fairly well - the mining thing may be a very abnormal use .
I have 250gb samsung 840 something C drive still going strong - use it a lot as a work space as well -to save time - Windows was probably paging to it , or what ever it does reasonably often

 

[Hysc84]

I think it is because of the user usage pattern, deemed it as storage and didnt perform trim maintenance as often

 

[stewi0001]

Based on this video that explains how SSD works, the main wear and tear that leads to failure is deleting.

 

[defaultluser]

Of course they're going to see the same trends - when you use consumer gear I the Datacenter, you should expect similar die-off rates(hdd vs SSD)

 

[Superconductor]

Well it is backblaze: even a tiny bit of research will let you find they do this kind of thing all the time: Take consumer hardware and then put it to worst-case-scenario data center loads: of course it's going to fail quicker and more often and it serves the company well since their business is predicated on selling you backup solutions.

You missed my point entirely