Western Digital Readies 28TB HDD

[darwin006]

Summary

Western Digital is set to begin sampling its 28TB nearline hard drive (HDD) designed for hyperscale datacenters in the coming weeks, the company revealed at its earnings call.

 

Quotes

Quote

 The HDD will employ the company's energy-assisted perpendicular magnetic recording (ePMR) and UltraSMR (shingled magnetic recording) technologies to deliver the unprecedented capacity.

 

"This cutting-edge product is built upon the success of our ePMR and UltraSMR technologies with features and reliability trusted by our customers worldwide. We are staging this product for quick qualification and ramp as demand improves."

 

An intriguing aspect of Western Digital's 28TB HDD is that it relies on the company's 2nd generation ePMR technology with refined write heads to enable higher areal density and thinner tracks. As Western Digital's UltraSMR technology boosts areal density of a conventional magnetic recording (CMR) media by about 20%, the company needs a 24TB HDD to build a 28TB UltraSMR drive.

 

Western Digital is currently shipping its 26TB UltraSMR hard drives, introduced over a year ago, to a select clients among hyperscalers. The qualification process for these UltraSMR drives naturally took some considerable time, due to the need for hyperscalers to understand the new technology's behavior and performance since UltraSMR uses a host of hardware, firmware, and software innovations

 

Western Digital's 28TB hard drives will contend with Seagate's 32TB hard drives, which utilize heat-assisted magnetic recording technology (HAMR). Seagate's product, which is being evaluated right now, is expected to ramp in early 2024, promising higher capacity and performance, particularly in write operations

 

Western Digital's HAMR-based hard drives are at least 1.5 years away

 

 

My thoughts

SSD manufactures past HDD manufactures in drive capacity a while ago but large capacity SSDs continue to be around 10x as much per TB

HDD manufactures continue to find ways to increase drive capacity and decrease the cost of bulk data storage.

HDDs may not be something that you have in your PC anymore but for bulk storage in a NAS or data center they dont seem to be going anywhere anytime soon.

 

Sources

https://www.tomshardware.com/news/western-digital-readies-28tb-hdd

[williamcll]

How fast does this go?

[TrigrH]

28 minutes ago, williamcll said:

How fast does this go?

The speed is not the real question:

Does it take longer to write the entire disk? yes.

 

[Kisai]

1 hour ago, williamcll said:

How fast does this go?

If it's anything like the existing Red/Red Pro's, the drives often exhaust the cache and sound like they are power-cycling frequently.

[treestain]

1 hour ago, TrigrH said:

The speed is not the real question:

Does it take longer to write the entire disk? yes.

 

I'd love to just imagine a large business using these drives and then 10 years down the road having to shred all of the drives for security reasons like holy crap

[SorryBella]

2 hours ago, williamcll said:

How fast does this go?

3 hours ago, darwin006 said:

UltraSMR

more like seconds per MB am i right fellas?

[Franck]

5 hours ago, williamcll said:

How fast does this go?

For the market this is not a useful metric.

Even if it would write 1000 time slower than another drive the reason why someone buy these is because this gets [x] TB for [y] cubic centimeters.

[darknessblade]

5 hours ago, TrigrH said:

The speed is not the real question:

Does it take longer to write the entire disk? yes.

 

With a 100mib connection, and nonstop downloading, it would take about a month to fill it up.

 

[X-System]

2 hours ago, darknessblade said:

With a 100mib connection, and nonstop downloading, it would take about a month to fill it up.

 

And for them who still use 56k modem (about 4.6 kbps), it would take 193 years

[Beskamir]

Resilvering an array with one of these dead is gonna be a nail biting few days during which another drive will be very likely to also fail.

[PocketNerd]

So HDDs are becoming the next Tape, cool.

[Godlygamer23]

On 8/3/2023 at 2:23 AM, Beskamir said:

Resilvering an array with one of these dead is gonna be a nail biting few days during which another drive will be very likely to also fail.

This is why I've been conservative with my drive capacities within my server. It becomes a slippery slope when you're trying to recreate data on a replacement drive since the transfer rates aren't going up at the same pace as the capacity. 

 

Basically, the drives are too big for their transfer speed capabilities, which during an array rebuild, results in a much higher chance of another drive failing. Now, one should have all their data backed up, but that creates another problem. 

[RejZoR]

On 8/2/2023 at 9:09 AM, treestain said:

I'd love to just imagine a large business using these drives and then 10 years down the road having to shred all of the drives for security reasons like holy crap

They shred them mechanically. They just throw them in the shredder and tiny bits come out the other end. Good luck recovering data from that. No one is going to do software wipe on them as they won't be worth anything anyway as empty drives. And if they fail before, they go into the same shredder anyway since you can't wipe them in software, but the platters still contain data.

[OhioYJ]

On 8/2/2023 at 1:39 AM, TrigrH said:

The speed is not the real question:

For me it is. I'd like a larger HD here at some point. My Steam Library for example currently resides on a 22 TB WD Gold drive (the limit of what they offer at the moment).

[leadeater]

On 8/3/2023 at 6:23 PM, Beskamir said:

Resilvering an array with one of these dead is gonna be a nail biting few days during which another drive will be very likely to also fail.

Basically nobody is going to be using these in a RAID like setup. They'll go in to software defined storage systems like Ceph or Gluster and be independent disks with either replica data copies or erasure coding data chunks.

 

Even ZFS is not the best tool for the job past many hundred TB of data.

 

These will be used in systems that utilize many servers and will be many hundreds to thousands of disks. Disk level failure domains are not what is done here, lowest commonly done is server/host level i.e. I use EC 8+5 across 15 servers which means 5 servers can go offline and data is still readable and recoverable (each server has 24 HDDs).

 

Anyone actually wanting to use them with ZFS needs to wait for the ePMR variant.

[StDragon]

Basically, you're not going to be running any VMs or IOPS sensitive operations on UltraSMR drives. They're primarily for object storage; specifically archival retention and in many cases immutable backups.

And to @leadeater point, fault-tolerance is handled at the node level than the disk. aka, RAIN (Redundant Array of Independent Nodes). But I'm old-school, I feel more comfortable working with and troubleshooting classic RAID volumes (even ZFS and Btrfs types). That level of scale-out abstractions has many advantages, but for home/personal use, I'll stick with RAID.

[jagdtigger]

On 8/3/2023 at 8:23 AM, Beskamir said:

Resilvering an array with one of these dead is gonna be a nail biting few days

Its smr so you can safely assume several weeks of resilver time.....

(There is a test on youtube, a hdd afraction of the size of this thing tok more than a week to reailver....)

[StDragon]

1 hour ago, Dean0919 said:

Same here. I tend to leave my games installed on even if I finished them and since I have 400 games on Steam, I'm still using HDD

With a 1 or 2 TB SSD (could get away with 512GB or smaller, but at current prices why skimp?), something like PrimoCache would pair nicely with a Steam library on the HDD. Games would get cached intelligently depending on how often they're accessed. Old cached data would be flushed out to make room for any new content requested off the HDD once the SSD is full; the caching manages itself algorithmically.

[AnonymousGuy]

On 8/8/2023 at 9:29 AM, Godlygamer23 said:

This is why I've been conservative with my drive capacities within my server. It becomes a slippery slope when you're trying to recreate data on a replacement drive since the transfer rates aren't going up at the same pace as the capacity. 

 

Basically, the drives are too big for their transfer speed capabilities, which during an array rebuild, results in a much higher chance of another drive failing. Now, one should have all their data backed up, but that creates another problem. 

I'm not particularly concerned about 3 drives failing within a few days of each other.  This was being talked about when 4TB drives were coming out "omg raid is dead it'll literally be impossible to rebuild" and here we are...

 

In any case drives are so cehap I replicate the entire thing on another set of shittier drives.

[Godlygamer23]

9 hours ago, AnonymousGuy said:

I'm not particularly concerned about 3 drives failing within a few days of each other.  This was being talked about when 4TB drives were coming out "omg raid is dead it'll literally be impossible to rebuild" and here we are...

 

In any case drives are so cehap I replicate the entire thing on another set of shittier drives.

4TB is quite a bit different from 28TB. Quite literally, 28TB is 7x bigger than 4TB, so even if you filled only half of the drive's capacity, that's still potentially double the amount of time, maybe more. Things get worse if you happen to store a bunch of small files, which is where archive files/folders can come into play. 

 

Drives might be cheap, but data isn't. So if you decide to backup your data with shit drives, and your main drives fail during an array rebuild, and you try to get that data back from your 'shitter' drives, then those could fail too depending on how shitty they really are, and how often you use them. I personally back up my data to a 12TB WD Red Plus, and then encrypt and upload my really important files to Dropbox. 

[AnonymousGuy]

On 8/12/2023 at 8:43 AM, Godlygamer23 said:

 Things get worse if you happen to store a bunch of small files, which is where archive files/folders can come into play. 

The rebuild has no awareness about the filesystem or contents.  It's all bits.

[TetraSky]

Rebuilding a RAID array with these in it would probably take a few weeks eh.

 

If only high capacity SSDs were readily available for much cheaper than they are right now... Just recently there was that announcement about a 256TB SSD. Too bad it will likely cost as much as a high end car.

[leadeater]

1 hour ago, AnonymousGuy said:

The rebuild has no awareness about the filesystem or contents.  It's all bits.

Not really, for good old RAID yes but for ZFS it does actually matter

 

Quote

ZFS uses variable-sized blocks. Therefore, for each recordsize wort of data, which can be anywhere from 4KB to 1MB, ZFS needs to consult the block pointer tree to see how data is laid out on disks. Because block pointer trees are often fragmented and files are often fragmented, there is quite a lot of head movement involved. Rotational hard drives perform much slower with a lot of head movement, so megabyte per second speed of the rebuild is slower than that of a traditional RAID. Now, ZFS only rebuilds the part of the array which is in use and it does not rebuild free space. Therefore, on lightly used pools it may actually complete faster than a traditional RAID. However, this advantage disappears as the pool fills up.

https://blocksandfiles.com/2022/06/20/resilvering/

[AnonymousGuy]

On 8/13/2023 at 4:55 PM, TetraSky said:

Rebuilding a RAID array with these in it would probably take a few weeks eh.

3 days.  That's 3x how long my 10TB drives take to rebuild.

On 8/13/2023 at 5:50 PM, leadeater said:

Not really, for good old RAID yes but for ZFS it does actually matter

 

https://blocksandfiles.com/2022/06/20/resilvering/

I'm a simple man running raid 6 with NTFS on top...and SMB (but hey at least we have dual 10Gb for SMB multithread...).  It's dumb and basic but it's all I really want to be bothered maintaining and setting up when I need the thing to run plex and a file host accesible to the world, and be accessible to other windows boxes.  And host 2 windows VMs.  And be available 24/7/365 where I can't take it down for a week to clean slate.  I'm in a mental state where I'm annoyed when I have to do SSL certificate renewal every year.

[leadeater]

1 hour ago, AnonymousGuy said:

I'm a simple man running raid 6 with NTFS on top...and SMB (but hey at least we have dual 10Gb for SMB multithread...).  It's dumb and basic but it's all I really want to be bothered maintaining and setting up when I need the thing to run plex and a file host accesible to the world, and be accessible to other windows boxes.  And host 2 windows VMs.  And be available 24/7/365 where I can't take it down for a week to clean slate.  I'm in a mental state where I'm annoyed when I have to do SSL certificate renewal every year.

I still love all my hardware RAID cards and use them for various things, hard to beat them in low HDD count arrays.