Desktop vs Enterprise HDD - Failure Rate Analysis. Do desktop hard drives really fail sooner?
ฝัง
- เผยแพร่เมื่อ 17 พ.ย. 2023
- Desktop drives have a much shorter warranty and they can be cheaper, but do they really fail sooner? I deep dive on AFR (annualized failure rate) analysis on 4Tb to 8Tb disks from Seagate, HGST and Western Digital to find out.
I compare a variety of 4Tb, 6Tb and 8Tb models, including Seagate Barracuda vs Exos and WD Deskstar vs Ultrastar and Megascale DC. Based on nearly 5 billion hours of data from 95,000 drives. Many of the drives have data for up to 9 years of use. You might be surprised what I find.
Video on the broader analysis of 430k drives over 10 years of data : • Comparing Seagate vs W...
Dive into AFR data on 10-16Tb Enterprise disks : • Failure Rate Analysis ...
Enterprise vs NAS disks? Which should you choose? : • Exos vs IronWolf Pro -...
Link to the BackBlaze data source : www.backblaze.com/cloud-stora...
You can support me at www.buymeacoffee.com/sometechguy
Thank you to everyone for watching! - วิทยาศาสตร์และเทคโนโลยี
I spent years working in aerospace and one of my first projects was to analyze drive failures as a way to justify migrating from HDD to SSD. We were dealing with high vibration applications so there were some significant variables that you're not accounting for.
I don't think most people realize how valuable this information actually is... much appreciated. Hardware failure rates are constantly being assessed and reassessed as part of warranty reduction. I know tech guys want the best drive they can get but the corporate overlords only care about long term cost reductions and that's all driven by failure rates divided by dollar signs amortized over the warranty coverage period.
The way you use stats and explain your analytical process for these videos is top notch!
Been looking at getting a NAS system, as a musician and stats nerd who shoots a lot of video. Your videos have been consistently awesome & helpful, so thank you so much!
PS: I’ve noticed your audio still has those digital clicks and pops, I can help troubleshoot the issue if you need
Thank you for the feedback, its really appreciated.
I have been trying to resolve the audio glitches, and not got to the bottom of the cause yet, so if you have any ideas then I would appreciate them, it doesn't seem to caused by external audio, such as plosives but seems to be in the encoding. You can get me though my channels contact email.
The stat if the WD drive is frightening considering how popular these are for NAS
I don't know if they got a bad batch, or what happened there. But yes, bad failure rates and from the data it looked like they all got decommissioned pretty early in their life.
This is a great video, thanks so much for taking the time to put this together! This is awesome😊
You are welcome. Glad you found it valuable and appreciate you taking the time to leave a comment.
From 35 years in IT, most of it in infrastructure especially storage, avoid the divisible by 3 capacities. 3, 6, 12 etc. They're not a true engineering evolution, just 2, 4, 8 etc drives where the manufacturer is pushing the limits of existing Technology.
I have not heard this before, and that's an interesting take!
The 5Tb disks always felt odd to me, and not sure why. But back in the day, disks were not so uniform and I remember building a PC with a 42Mb Conner IDE disk.
My powers of 2 bias really was justified LOL
With that logic how about 16, 18, 20 and 22 TB all on practically same tech, just small incremental platter improvements.
I do remember early 3tb drives being really problematic across brands IME, but ofc that's just my anecdote.
My 2x hgst 6tb just running fine since 2015
I have two 22 TB WD Gold in NAS (in RAID 1, I’m planning to add two more, and go RAID 5 eventually). 6 months of 24/7 - so far so good. I bought them for long term data storage - will see how it plays out. I have WD Scorpio Black 750 GB laptop HDD from 2011., still functioning perfectly fine. Also, got numerous external portable drives in the last 5 years from WD/Seagate (2-5 TB ones), all working just fine. I am interested in their reliability, as I had a bad experience with a desktop drive some 20 years ago (don’t remember what it was, only capacity of 40 GB), that tought me to have backups for everything. But in general, I’ve found all spinners very reliable for long term storage, that’s why I went for Enterprise WD’s (price for 22 tb Gold's was good at the time). So far, no buyer’s remorse, just wondering if Seagate is a smarter choice these days. Great data analysis here, looking forward for more data on these.
The number one killer for hard drives is HEAT. Keep the heat down "proper cooling" and the drive will last a long time.
On that note, drives will fail, any mechanical device will eventually fail over time, depending how long and heavily it is being utilized.
Great work
Very helpful video! Which stock footage are you using?
Thanks for this. I just installed (
I have a Seagate ATA internal drive that was bought in 2003. Still running. I also have a Seagate Freeagent 3TB drive, dead on the week after warranty expired. Now all my data are mirrored
Isn't the smaller consumer seagate disks smr drives?
We also don't know how backblaze are using the drives.
Are they mainly "hot storage" or is there frequent writes.
Smr drives probably scream for mercy in a enterprise setup with frequent writes.
Since the usage of these drives are not known it's difficult to judge without how they are used.
Or if they are run in raid setups or zfs etc..
No comments on data integrity? Failure rate is one thing, but enterprise drives normally advertise (per their data sheets) a lower UBER than consumer-grade drives, usually by a factor of ten.
This is a good point, I need to go look at the data and see how well unrecoverable bit errors are recorded in SMART to see if desktop drives really suffer from a higher bit error rate or not. It seems common for SMART to report numbers for Raw read error rates, but record pretty much zero unrecoverable errors. But this may depend on the workload and how often data is actually read, for example if scrubs are performed.
If the data looks useful, I could incorporate that in future comparisons.
Cool, Hitachi and IBM hardisks were the most durable, I hope they get revived, though they're maybe a bit more expensive (just a bit) which may not be for enterprises but it's perfect for homes.
God bless.
I have a bit over half of my WD60EFRX Red plus die between 3 and 5 years. Will se how my WD80EFAX red plus will perform over time ... new ones might be a different brand, as prices now are much cheaper for the Toshiba N300 and then Seagate Ironwolf NAS.
The Backblaze data contains very little WD info until recently, where they started deploying larger Ultrastar disks that came from the HGST aquisition. But I personally moved from WD Reds, first to Ironwolf Pros, but I would take a look at my video on NAS vs Enterprise disks, as I mostly buy Exos Enterprise disks recently, as the price point, reliability and warranty seem like a a great choice. At least, this has been true recently in places like the US and the UK as well as Germany. Good luck with whatever you choose next!
Ordered Exos x18 18TB for a HDD upgrade, current price was 14€ / TB cheapest of all.
How often should a HDD be powered up to "keep it healthy"? I am using it as cold storage and am trying to figure out the best type of drives to use.
An HDD will be far better than an SSD for this. Data on SSD degrades naturally, where this doesn’t happen on a hard disk. Theoretically the data should last a very long time, but bitrot can occur due to external factors, such as solar storms / cosmic rays but I would say this isn’t usually a serious issue. Powering up a disk alone may not find those and you might actually need to read the data to identify errors, which may or may not be correctable.
If it’s critical data, keep more than one copy and then you are extremely unlikely to have a problem.
@@sometechguy ok thank you.
Don't know about the bearings in todays harddrives.
Remember the earlier fluid bearing discs may have issues when they starte to get older.
The platter "lock up" but you could twist the discs hard to loose the platters inside.
(happend with me on two discs but that was with drives from around 2010'ish era)
Technology probably got a lot better today, if you already not using wery old drives for this..
I'm surprised that WD Red is CMR because I thought all Reds were those bad SMR. SMR would explain the graph, and it's a bit weird how WD suddenly forgot how to make CMR drives reliable.
Red drives used to be CMR, and then they covertly changed them to SMR without changing the model numbers in any significant way and without stating it on the datasheet.
Then, following backlash, they eventually stated that they were in fact now SMR and created a new line called Red Plus, which were really just the old Red drives and were again CMR.
The 'EFRX' model in this analysis is the older CMR version, which was replaced by the 'EFAX' version, which was SMR.
The fact this is so confusing highlights why this was such a bad thing to do for customers. 🤔
Could you make your visualization available to us? I'd like to poke at this myself but i dont know how to visualize their data myself :(
Not easily currently. But I may create a site where I can more easily share data and written content also. But if you want to play around and create different visuals, you may need to work with the raw data.
@@sometechguy Alright, thanks!
Funny thing, one of my 6tb drives failed, a wd red.
Funny, but also not funny at the same time. 😢
@@sometechguy and now the second one starts failing, right when I'm rebuilding the raid pool
@@PetervanderKruys I had the same thing happen with some Red 6Tb. They were all at around 7 years, I had one fail and had one other fail during rebuild. RAID rebuilds are one of the most stressful things you can do to disks in the array, so if any other are marginal, its the time they are likely to fail. Before doing rebuilds, I would recommend validating backups look solid. And RAID 6 or other arrays with N+2 redundancy can mitigate this a little but the more disks in the array, the more likely this is to happen.
Hope it isn't catastrophic for you, sometimes the other disk just reports some errors and maybe minimal files are impacted. And hopefully anything important also available on a backup. Good luck!
@@sometechguy yeah thanks, I hope all is good. Fortunately it is only 1 bad sector that popped up, but that’s how the initial drive started. After 3 bad sectors it got some initialisation problems but once I took it out and put it back, my synology just wouldn’t accept it. The raid was degraded and stayed that way.
I’m using SHR from my Synology, so advanced raid5 with a hot spare (which unfortunately was a bit smaller than the failed drive, so useless). I can use that drive to replace the smaller failing drive and get another IronWolf 6TB as hot spare. I tried an IronWolf Pro 8TB but either it was broken or just didn’t work on my old Synology.
But damn I wished I had a raid6 variant right now, food for thought!
Spinning.
Can this be shortcut place to get a job about harddisk?
Planning to create social network and community network, 2 different website. And also planning to create webhosting place too. Life imagination of like this was fun, then like when speaking of it they recreate an idea and like a publishment about it to my eyes. Other business is c.s 1.6 servers.
SEAGATE ALWAYS SUCKS!! ALWAYS! IN MY OERSONAL LIFE AND IN ANY STATS
Could say the same of WD from my experience on multiple units. Unfortunately speaking about drives seems there's no thing as a safe bet.