“The two models, the 30TB … and the 32TB …, each offer a minimum of 3TB per disk”. Well, yes, I would hope something advertised as being 30TB would offer at least 3TB. Am I misreading this sentence somehow?
They probably mean the hard drive has 10 platters, each containing at least 3TB.
Haven’t bought Seagate in 15 years. They improve their longevity?
https://www.backblaze.com/cloud-storage/resources/hard-drive-test-data
Backblaze reports are cool
I have one Seagate drive. It’s a 500 GB that came in my 2006 Dell Dimension E510 running XP Media Center. When that died in 2011, I put it in my custom build. It ran until probably 2014, when suddenly I was having issues booting and I got a fresh WD 1 TB. Put it in a box, and kept it for some reason. Fast forward to 2022, I got another Dell E510 with only an 80 GB. Dusted off the old 500 GB and popped it in. Back with XP Media Center. The cycle is complete. That drive is still noisy as fuck.
Not worth the risk for me to find out lol. My granddaddy stored his data on WD drives and his daddy before him, and my daddy after him. Now I store my data on WD drives and my son will to one day. Such is life.
And here I am with HGST drives hitting 50k hours
Edit: no one ever discusses the Backblaze reliability statistics. Its interesting to see how they stack up against the anecdotes.
Vastly. I’m running all seagate ironwolf pros. Best drives Ive ever used.
Used to be WD all the way.
whoa
Dude
Seagate. The company that sold me an HDD which broke down two days after the warranty expired.
No thanks.
laughing in Western Digital HDD running for about 10 years nowWestern digital so good
Had the same experience and opinion for years, they do fine on Backblaze’s drive stats but don’t know that I’ll ever super trust them just 'cus.
That said, the current home server has a mix of drives from different manufacturers including seagate to hopefully mitigate the chances that more than one fails at a time.
I currently have an 8 year old Seagate external 4TB drive. Should I be concerned?
Funny because I have a box of Seagate consumer drives recovered from systems going to recycling that just won’t quit. And my experience with WD drives is the same as your experience with Seagate.
Edit: now that I think about it, my WD experience is from many years ago. But the Seagate drives I have are not new either.
Survivorship bias. Obviously the ones that survived their users long enough to go to recycling would last longer than those that crap out right away and need to be replaced before the end of the life of the whole system.
I mean, obviously the whole thing is biased, if objective stats state that neither is particularly more prone to failure than the other, it’s just people who used a different brand once and had it fail. Which happens sometimes.
Ah I wasn’t thinking about that. I got the scrappy spinny bois.
I’m fairly sure me and my friends had a bad batch of Western digitals too.
Did you buy consumer Barracuda?
I had the opposite experience. My Seagates have been running for over a decade now. The one time I went with Western Digital, both drives crapped out in a few years.
I have 10 year old WDs and 8 year old Seagates still kicking. Depends on the year. Some years one is better than others.
Heck yeah.
Always a fan of more storage. Speed isn’t everything!
HP servers have more fans!
Good. However, 2 x 16TB Seagate HDDs still cheaper, isn’t it?
These drives aren’t for people who care how much they cost, they’re for people who have a server with 16 drive bays and need to double the amount of storage they had in them.
(Enterprise gear is neat: it doesn’t matter what it costs, someone will pay whatever you ask because someone somewhere desperately needs to replace 16tb drives with 32tb ones.)
In addition to needing to fit it into the gear you have on hand, you may also have limitations in rack space (the data center you’re in may literally be full), or your power budget.
How many platters?!
30 to 32 platters. You can write a file on the edge and watch it as it speeds back to the future!
The two models, […] each offer a minimum of 3TB per disk
Huh? The hell is this supposed to mean? Are they talking about the internal platters?
More than likely
It never ceases to amaze me how far we can still take a piece of technology that was invented in the 50s.
Talking about steam, steam-powered things are 2 thousand years old at least and we still use the technology when we crack atoms to make energy.
What the Romans had wasn’t comparable with an industrial steam engine. The working principle of steam pushing against a cylinder was similar, but they lacked the tools and metallurgy to build a steam cauldron that could be pressurized, so their steam engine could only do parlor tricks like opening a temple door once, and not perform real continuous work.
This isn’t unique to computing.
Just about all of the products and technology we see are the results of generations of innovations and improvements.
Look at the automobile, for example. It’s really shaped my view of the significance of new industries; we could be stuck with them for the rest of human history.
That’s how most technology is:
- combustion engines - early 1900s, earlier if you count steam engines
- missiles - 13th century China, gunpowder was much earlier
- wind energy - windmills appeared in the 9th century, potentially as early as the 4th
Almost everything we have today is due to incremental improvements from something much older.
Solid state is kinda like a microscopic punch card.
So are optical discs
Much more so than solid state.
More like microscopic fidget bubble poppers.
When the computer wants a bit to be a 1, it pops it down. When it wants it to be a 0, it pops it up.
If it were like a punch card, it couldn’t be rewritten as writing to it would permanently damage the disc. A CD-RW is basically a microscopic punch card though, because the laser actually burns away material to write the data to the CD.
They work through electron tunneling through a semiconductor, so something does go through them like an old punch card reader
Current ones also store multiple charge levels per cell, so they’re no longer one bit each. They have multiple levels of “punch” for what used to just be one bit.
Just a reminder: These massive drives are really more a “budget” version of a proper tape backup system. The fundamental physics of a spinning disc mean that these aren’t a good solution for rapid seeking of specific sectors to read and write and so forth.
So a decent choice for the big machine you backup all your VMs to in a corporate environment. Not a great solution for all the anime you totally legally obtained on Yahoo.
Not sure if the general advice has changed, but you are still looking for a sweet spot in the 8-12 TB range for a home NAS where you expect to regularly access and update a large number of small files rather than a few massive ones.
Oh hey, I did something right. That’s kinda neat
I am troubled in my heart. I would not have been told so in this way.
honestly curious, why the hell was this downvoted? I work in this space and I thought this was still the generally accepted advice?
Because people are thinking through specific niche use cases coupled with “Well it works for me and I never do anything ‘wrong’”.
I’ll definitely admit that I made the mistake of trying to have a bit of fun when talking about something that triggers the dunning kruger effect. But people SHOULD be aware of how different use patterns impacts performance, how that performance impacts users, and generally how different use patterns impact wear and tear of the drive.
Come on man, everything, and mean everything you said is wrong.
Budget tape backup?
No, you can’t even begin to compare drives to tape. They’re completely different use cases. A hard drive can contain a backup but it’s not physically robust to be unplugged, rotated off site , and put into long term storage like tape. You might as well say a Honda Accord is a budget Semi tractor trailer.
Then you specifically called out personal downloads of anime as a bad use case. That’s absolutely wrong in all cases.
It is absurd to imply that everyone else except for you is less knowledgeable and using a niche case except you.
Because everything he said was wrong?
Not a great solution for all the anime you totally legally obtained on Yahoo.
Mainly because of that. Spinning rust drives are perfect for large media libraries.
There isn’t a hard drive made in the last 15 years that couldn’t handle watching media files. Even the SMR crap the manufacturers introduced a while back could do that without issue. For 4k video you’re going to see average transfer speeds of 50MB/s and peak in the low 100MB/s range, and that’s for high quality videos. Write speed is irrelevant for media consumption, and unless your hard drive is ridiculously fragmented, seek speed is also irrelevant. Even an old 5400 RPM SATA drive is going to be able to handle that load 99.99% of the time. And anything lower than 4K video is a slam dunk.
Everything I just said goes right out the window for a multi-user system that’s streaming multiple media files concurrently, but the vast majority of people never need to worry about that.
The fundamental physics of a spinning disc mean that these aren’t a good solution for rapid seeking of specific sectors to read and write and so forth.
It’s no ssd but is no slower than any other 12TB drive. It’s not shingled but HAMR. The sectors are closer together so it has even better seeking speed than a regular 12TB drive.
Not a great solution for all the anime you totally legally obtained on Yahoo.
???
It’s absolutely perfect for that. Even if it was shingled tech, that only slows write speeds. Unless you are editing your own video, write seek times are irrelevant. For media playback use only consistent read speed matters. Not even read seek matters except in extreme conditions like comparing tape seek to drive seek. You cannot measure 10 ms difference between clicking a video and it starting to play because of all the other delays caused by media streaming over a network.
But that’s not even relevant because these have faster read seeking than older drives because sectors are closer together.
So I’m guessing you don’t really know what you’re talking about.
Not sure what you’re going on about here. Even these discs have plenty of performance for read/wrote ops for rarely written data like media. They have the same ability to be used by error checking filesystems like zfs or btrfs, and can be used in raid arrays, which add redundancy for disc failure.
The only negatives of large drives in home media arrays is the cost, slightly higher idle power usage, and the resilvering time on replacing a bad disc in an array. Your 8-12TB recommendation already has most of these negatives. Adding more space per disc is just scaling them linearly.
Additionally, most media is read in a contiguous scan. Streaming media is very much not random access.
Your typical access pattern is going to be seeking to a chunk, reading a few megabytes of data in a row for the streaming application to buffer, and then moving on. The ~10ms of access time at the start are next to irrelevant. Particularly when you consider that the OS has likely observed that you have unutilized RAM and loads the entire file into the memory cache to bypass the hard drive entirely.
I’m real curious why you say that. I’ve been designing systems with high IOPS data center application requirements for decades so I know enterprise storage pretty well. These drives would cause zero issues for anyone storing and watching their media collection with them.
HDD read rates are way faster than media playback rates, and seek times are just about irrelevant in that use case. Spinning rust is fine for media storage. It’s boot drives, VM/container storage, etc, that you would want to have on an SSD instead of the big HDD.
Just one would be a great backup, but I’m not ready to run a server with 30TB drives.
I’m here for it. The 8 disc server is normally a great form factor for size, data density and redundancy with raid6/raidz2.
This would net around 180TB in that form factor. Thats would go a long way for a long while.
I dunno if you would want to run raidz2 with disks this large. The resilver times would be absolutely bazonkers, I think. I have 24 TB drives in my server and run mirrored vdevs because the chances of one of those drives failing during a raidz2 resilver is just too high. I can’t imagine what it’d be like with 30 TB disks.
Yeah I agree. I just got 20tb in mine. Decided to just z2, which in my case should be fine. But was contemplating the same thing. Going to have to start doing z2 with 3 drives in each vdev lol.
A few years ago I had a 12 disk RAID6 array and the power distributor (the bit between the redundant PSUs and the rest of the system) went and took 5 drives with them, lost everything on there. Backup is absolutely essential but if you can’t do that for some reason at least use RAID1 where you only lose part of your data if you lose more than 2 drives.
Is RAID2 ever the right choice? Honestly, I don’t touch anything outside of 0, 1, 5, 6, and 10.
Edit: missed the z, my bad. Raidz2 is fine.
raidz2 is analogous to RAID 6. It’s just the ZFS term for double parity redundancy.
I can’t wait for datacenters to decommission these so I can actually afford an array of them on the second-hand market.
Home Petabyte Project here I come (in like 3-5 years 😅)
better start preparing with a 10G network!
Way ahead of you… I have a Brocade ICX6650 waiting to be racked up once I’m not limited to just the single 15A circuit my rack runs off of currently 😅
Hopefully 40G interconnect between it and the main switch everything using now will be enough for the storage nodes and the storage network/VLAN.
Exactly, my nas is currently made up of decommissioned 18tb exos. Great deal and I can usually still get them rma’d the handful of times they fail
Where is a good place to search for decommissioned ones?
eBay sellers that have tons of sales and specialize. You can learn to read between the lines and see that decom goods are what they do.
SaveMyServer is a perfect example. Don’t know if they sell drives though.
Serverpartdeals has done me well, drives often come new enough that they still have a decent amount of manufacturers warranty remaining (exos is 5yr) and depending on the drive you buy from them spd will rma a drive for 5 years from purchase (but not always, depends on the listing, read the fine print).
I have gotten 2 bad drives from them out of 18 over 5 years or so. Both bad drives were found almost immediately with basic maintenance steps prior to adding to the array (zeroing out the drives, badblocks) and both were rma’d by seagate within 3-5 days because they were still within the mfr warranty.
If you’re running a gigantic raid array like me (288tb and counting!) it would be wise to recognize that rotational hard drives are doomed and you need a robust backup solution that can handle gigantic amounts of data long term. I have a tape drive for that because I got it cheap at an electronics recycler sold as not working (thankfully it was an easy fix) but this is typically a super expensive route. If you only have like 20tb then you can look into stuff like cloud services, bluray, redundant hard drive, etc. or do like I did in the beginning and just accept that your pirated anime collection might go poof one day lol
What kind of tape drive are you using? My array isn’t as large as yours (120tb physical), but it’s big enough that my only real options for backup are tape or a whole secondary array for just backup.
Based on what I’ve seen, my options are a prohibitively large number tapes with an older LTO standard or prohibitively expensive tapes with a newer LTO standard.
My current backup strategy consists of automated backups to Backblaze B2 for the really important stuff like personal documents or projects and hoping my ZFS array doesn’t fail for everything else.
I have an ibm qualstar lto8 drive. I got it because I gambled, it was cheap because it was throwing an error (I forget what the number was) but it was one that indicates an issue in the tape path. I was able to get the price to $150 because I was buying some other stuff and because ultimately if the head was toast it was basically useless. But I got lucky and cleaning the head and tape path brought it back to life. Dunno how long it will last. I’ll live with it though because buying one that’s confirmed working can be thousands
You’re right that lto8 tapes are pricey but they’re quite a bit cheaper than building an equivalent array for backup that is significantly more reliable long term. A tape is about 12tb and $40-50, although sometimes they pop up cheaper. I generally don’t back up stuff continually with this method, I back up newer files that haven’t been synced to tape once every six weeks or so. It’s also something that you can buy a bit at a time to soften the financial blow of course. Maybe if you get a fancy carousel drive you’d want to fill it up but frankly that just seems like it would break much easier
More modern tapes have support for ltfs and I can basically use it like an external hard drive that way. So it’s pretty much I pop a tape in, once a week or so I sync new files to said tape, then as it gets full I swap it for a new tape. Towards the end I print a directory of what’s on it because admittedly doing it this way is messy. But my intention with this is to back up my “medium critical” files. Stuff that if I lost I would be frustrated over, but not heartbroken. Movies and TV shows that I did custom muxes of to have my ideal subtitles, audio tracks, etc. all my dockers so stuff like my Jellyfin watch status and komga library stay intact, stuff like that. That takes up the bulk of my nas and my primary concerns are either the array fully failing or significant bit rot, and if either of those occur I would rebuild from scratch and just copy all the tapes back over anyway so the messy filing isn’t really a huge issue.
I also do sometimes make it a point to copy harder to find files onto at least 2 tapes on the outside chance a tape goes bad. It’s unlikely given I only buy new tapes and store them properly (I even go to the effort to store them offsite just in case my house burns down) but you never know I suppose
The advertised values of tape capacity is crap for this use. You’ll see like lto 8 has a native capacity of 12tb but a compressed capacity of 30tb per disk! And the disks will frequently just say 30tb on them. That’s nonsense here. Maybe for a more typical server environment where they’re storing databases and text files and shit but compressed movies and music? Not so much. I get some advantage because I keep most of my stuff in archival quality (remux/flac/etc) but even then I still usually dont get anywhere near 30tb
It’s pretty slow. Not the end of the world but just something to keep in mind. Lto8 is supposed to be 360MBps for uncompressed and 750MBps for compressed data but I don’t seem to hit those speeds at all. I’m not really in a rush though and everything verifies fine and works after copying back over so I’m not too worried. But it can take like 10-14 hours to fill a tape. If I ever do have to rebuild the array it will take AGES
For my “absolutely priceless” data I have other more robust backup solutions that are basically the same as yours (literally down to using backblaze, ha).
You got an incredible deal on your tape drive. For LTO8 drives, I’m seeing “for parts only” drives sold for around $500. I’d be willing to throw away $100 or $200 on the possibility that I could repair a drive; $500 is a bit too much. It looks like LTO6 is more around what my budget would be.; it would require a much larger number of tapes, but not excessively so.
I remember when BD-R was a reasonable solution for backup. There’s no way that’s true now. It really seems like hard drive capacity has far outpaced removable media. If most people are streaming everything, those of us who actually want to save their data locally are really the minority these days. There’s just not as much of a compelling reason for companies to develop cheap high-capacity removable discs.
I’m sure I’ll invest in a tape backup solution eventually, but for now, at least I have ZFS with paranoid RAIDZ.
Nice, where do you get yours?
also curious, buying new is getting too pricey for me
I personally use goharddrive and serverpartdeals on eBay and have had good luck, but I’m always looking for others
Never used goharddrive but can def endorse spd
This is for cold and archival storage right?
I couldn’t imagine seek times on any disk that large. Or rebuild times…yikes.
Random access times are probably similar to smaller drives but writing the whole drive is going to be slow
up your block size bro 💪 get them plates stacking 128KB+ a write and watch your throughput gains max out 🏋️ all the ladies will be like🙋♀️. Especially if you get those reps sequentially it’s like hitting the juice 💉 for your transfer speeds.
This is my favorite post ever.
Definitely not for either of those. Can get way better density from magnetic tape.
They say they got the increased capacity by increasing storage density, so the head shouldn’t have to move much further to read data.
You’ll get further putting a cache drive in front of your HDD regardless, so it’s vaguely moot.
For a full 32GB at the max sustained speed(275MB/s), 32ish hours to transfer a full amount, 36 if you assume 250MB/s the whole run. Probably optimistic. CPU overhead could slow that down in a rebuild. That said in a RAID5 of 5 disks, that is a transfer speed of about 1GB/s if you assume not getting close to the max transfer rate. For a small business or home NAS that would be plenty unless you are running greater than 10GiBit ethernet.
radarr goes brrrrrr
sonarr goes brrrrrr…
barrrr?
…dum tss!
Here i am still rocking 6TB.