> Unlike SSHDs, the OptiNAND drives do not store any user data at all during normal operation. Instead, the NAND is being used to store metadata from HDD operation in order to improve capacity, performance, and reliability.
This is done on zfs and (in some proprietary cases) in btrfs. It's interesting to see the concept implemented at a lower level. More tools to work with.
Ah yeah probably. HDDs have been doing lots of bookkeeping for a long time. I wonder if this change (from what?) actually provides any improvement to HDD performance.
Exactly. How is this any different from hybrid hard drives (10+ years ago)? https://www.tomshardware.com/reviews/seagate-momen... "Hybrid hard drives (H-HDDs) arrived in 2007, and they disappeared soon after because there was hardly any real benefit."
It's not clear what exactly is the metadata information is being stored here.
Also, rather than using actuator heads, why aren't drive heads a single linear piece solid-state hardware that has thousands of electronically addressable read/write heads on it?
Interesting idea for reducing latency, because a rotating disk would bring magnetic information towards (mostly) each solid-state read-write-head during one rotation, if placed tangentially? Saving another moving part within hdd's? Did not find stats for production cost of parts of hdds listed separately (also endurance and failure rate of hdd parts) or percentage of total cost of devices?
at least, the actuator would have reduced movement and acceleration latencies, if placed radially, for fine tuning between (maybe more expensive) read-write-heads?
One 22TB drive could hold pretty much the entire NTSC library of console videogames including Xbox 360/PS3. That will be fun for emulation. PS4/XBO+ and PC are much bigger, but still reasonable numbers.
Two drives could hold 44,000 movies in 1080p. There are fewer than 44,000 movies with more than 1000 votes on IMDb.
Most popular digital media will fit inside 300TB = 14 drives. Best case $4500 in today's money. 600TB would be even better. Music, TV, comics, books, and the text internet would be good collections too.
sheh is correct, 1GB is far too low an estimate for standard length (90-120 minute) HD movies encoded without excessive compression. 10GB may be a little high but calculation estimates of this sort should always be done conservatively by providing a little extra buffer overhead, so 10GB is much better to use as a a per-media-file size estimate than 1GB.
There's still plenty of Europe and Japan exclusive games however. And I have OCD and need to store all of them, so 22TB is nowhere near enough for the PS3 set alone :(
Don't forget to implement multi-Actuator on a single Actuator Arm and Dual Actuators in the same 3.5" body.
I see enough empty volume that with some re working, you could fit in 2x Actuator Arm Stacks with multi-actuators per arm leading to huge Linear R/W performance increases.
Dual actuators are already in production in some of Seagate's drives. I think all vendors will eventually adopt the scheme in order to scale sequential bandwidth / IOPS with capacity. For now, WD believes it is not cost-effective to implement dual actuators outside their R&D labs.
WD does have the triple-stage actuator in their single arm that enables finer positioning (which in turn means tracks can be bunched even closer together) in order to increase areal density.
Maybe I'm missing something, but we should be going back to 5.25" as it gives a large (50%?) surface area boost per platter, and the Google hard drive study suggested thicker drives with many more platters per drive. With this we're already looking at 40TB drives easily. Lowering rotational speed is no big deal.
A quick Amazon search yields a fair number of cases that feature 5.25" bays. I must be very lucky. I'd better go buy lottery tickets or something given my luck is so good.
So, will the iNAND solve the SMR delayed write problems? I don't claim this article said anything about it. I'm curious, because that's the major problem with SMR and if the heads had to move less because you have the metadata in iNAND that would at least help things.
The article mentions potential SMR capacity improvements, but doesn't mention performance. My guess is that it would provide a negligible boost. Once your drive fills up and your write cache fills up, every random write still requires reading several tracks + writing back several tracks. I'm not sure how many consecutive tracks are shingled in each shingled region... but let's say 5 for an example, so 10 revolutions per write at 7200rpm is ~83ms per write or 12 IOPS. Double the tracks per shingled region to get more capacity benefit from the shingling and you half the random write IOPS. A larger write cache might help you get in a few more performance writes before you hit the read-modify-write wall, but you'll still hit it eventually. SMR is pretty much not an option if you have a random write heavy workload. If any of my assumptions are flawed, please let me know.
The only flaw would be to put a random write heavy workload on HDDs these days: that's become "you're using it wrong".
The whole SMR debate really mostly centers around the RAID/ZFS rebuilds where you couldn't really avoid it and the upper layers (controller and FS) didn't know about SMR.
If they were SMR aware, rebuilds might perhaps be done at shingle granularity and here big flash (perhaps better outside the drive) would help.
Hardly any storage is designed to support constant maximum write pressure, because there isn't much you can save with optimizations. NAND could mostly just push the first level "SLC cache" for SMR much further so the SMR write amplification becomes less noticeable in "average scenarios".
If they use NAND to implement the fulll SMR write buffers, they better have lots of really good NAND. And I'd probably want a fail-back mode, where exhausted NAND just gets bypassed and won't result in a complete drive failure.
abufrejoval, you mention 2.5" 8TB CMR HDD. Yes, that would be lovely and I'd be an immediate buyer! FYI, I have just received email confirmation from WD, here in Australia, that their 4TB and 5TB Passport drives are SMR. Getting this confirmation from a senior tech support guy was like getting blood from a stone. I went through 4 stages of attempts with different levels of techs and a dozen emails back/forth. This fonal guy was rather matter of fact, clear and concise and confirmed it completely. This promptly put a stop to my planned purchase of four of the 5TB models. Just so you know!
The key role for this embedded NAND is to assist, and possibly actively tune, the basic mechanical challenge of hitting an impossibly small target on rotating platters:
> repeatable run out (RRO) recording of the head jitter / error position as the spindle revolves. This data (running into multiple gigabytes) is generated in the factory during manufacturing. It is typically stored in the disk, taking up space that could have potentially been used for user data. The OptiNAND architecture moves this to the NAND in the EFD.
Discussed in the slides here, at around the third or fourth slide.
Apparently, the mechanical “map” of each platter takes gigabytes of data, particular to each hard disk, each platter and segment. Although manufacturing tolerances on these things are amazing, each physical device has its bumps… to say nothing of non-uniform (at this scale) heat explanation.
The second role for this NAND is to buffer adjacent sectors (on nearby tracks as well as the same track) after writing… I didn’t realize how we’ve hit the wall there, having to re-write entire neighborhood of sectors every 6 writes or so:
> One of the downsides to increasing areal density by increasing the TPI is the need to do more frequent refreshes. From refreshing once in 10000 write operations in early HDDs, the narrow tracks now need to be refreshed as frequently as once every 6 writes.
> The second role for this NAND is to buffer adjacent sectors
No, I'm pretty sure it doesn't actually buffer data to be written. I think it's probably used to keep per-sector write statistics, so the drive can allocate sectors more intelligently and knows when to schedule rewrites.
Fascinating insights into how HDDs are transforming!
With rewrites of adjacent tracks dropping to 1 in 6, the real difference between CMR and SMR really starts to disappear.
The whole question as to where you'd want to put the intelligence still doesn't have a good answer: Here it's the drive, the smart RAID controller was in favor for a long time and I guess hyperscalers might just go for a (specialized?) server anyway, up to the point where the rewriting of shingles it a decision made by the host not the drive.
I'm surprised flash is judged fast enough for the operational data, but then again, the proper parts can be picked up in all those milliseconds that it take the heads to move/switch or the sectors to rotate: you can do a lot of compute in milliseconds these days!
For my personal use at these capacity points the RAID6/ZFS2 rebuild debate becomes a bit mote and a mirror set, for all its wasted capacity becomes the maximum that makes any sense. SMR would compensate that a little and I guess it's just time to complete the mental switch: HDD is tape and SSD is HDD.
I still wouldn't mind seeing 2,5" CMR drives with 8TB, but I guess that species has gone extinct.
I use 2.5" drives for cheap non volatile raid for booting desktops. All they really good for. WD has a single platter 1tb super thin, draws less currrent than ssd, 128mb cache. They fit anywhere in a PC case.
> With rewrites of adjacent tracks dropping to 1 in 6, > the real difference between CMR and SMR really starts to disappear.
Not really. If the controller allocates sectors so they're roughly grouped by the number of writes, then you're only rewriting a zone 1/6th of the time. That's still way less than *every* time.
abufrejoval, you mention 2.5" 8TB CMR HDD. Yes, that would be lovely and I'd be an immediate buyer! FYI, I have just received email confirmation from WD, here in Australia, that their 4TB and 5TB Passport drives are SMR. Getting this confirmation from a senior tech support guy was like getting blood from a stone. I went through 4 stages of attempts with different levels of techs and a dozen emails back/forth. This fonal guy was rather matter of fact, clear and concise and confirmed it completely. This promptly put a stop to my planned purchase of four of the 5TB models. Just so you know!
I've just purchased an 18TB WD GOLD HDD and have a few 14TB Ultrastars in my home server setup. My opinion is that they are simply the very best HDDs that have ever existed on the planet. I expect I will have these same drive a decade or two into the future - they are SO well made. Let's hope the helium stays contained eh?! I do a lot of 4K video editing. I bought my 18TB GOLD due to the fact all the reviews and WD itself said that was the biggest CMR drives could ever get. The 20TB Ultrastar was, of course, SMR. So NOW they come out and say 20TB will indeed be CMR too!! I don't know about you but it all sounds a bit 'too' experimental if they are now able to cram an extra 2TB onto the 18TB CMR disk simply because they have a 64GB slither of flash on the side! If I am missing anything I would really appreciate being brought up to speed on this. Cheers all - oh, my 1st post here tonight! :-)
The top-capacity HDDs (irrespective of the vendor) for the last few years are quite reliable because the platforms they are based on have been created mainly for enterprise customers. Unless those pass enterprise / cloud-deployment qualification, they do not make it out to the other product lines - simply because the selling price is too high for non-power users to adopt them, and there is no incentive for the vendors to face a backlash (likely that top capacity sales are not a big chunk of the revenue from the consumer segment).
Meh. Lipstick on a pig. Both, the product and thinly veiled advert, masked as an article. Where are substantial improvements ? Like multiple headstacks, with multiple active heads within each stack ? Also, capacity jump is minimal. Just 2 extra TB and those are gained by losing signal integrity, because they are pushing data density on the same process.
They are also cooking the public to gradually accept SMR through adjacent track degradation trick. By squeezing tracks ever more closer, far more thatn what the haeds can reliably adress, they are squeezing SMR into all drives.
"Where are substantial improvements ? Like multiple headstacks, with multiple active heads within each stack ?"
Those kind of improvements will require improvements like the ones detailed to even be feasible. Even then, they're still non-trivial, and they're going to add significantly to costs.
Plenty of people *are* putting a lot of time, energy, & money into higher-capacity NAND. However, HDDs still offer more GB/$ and modern, high-capacity NAND cannot be used for cold storage.
I agree about the WD golds. I thought 12tb was perfect price/size and these drives are phenomenal. Fast, too. Faster than red 12tb drives, or esp the white label 12tb drives they use in their externals. I am constructing a NAS, and raid 01 or 10 seems the way to go for 24gb of storage. I think raid 5 is not used as often as the past.
I have tried to hold onto all my data going back to the 90s. I have been through this a few times. I have to have two copies at least other than live. Older drives come in handy, as the 2nd copy can be jbod, but first copy of the live storage needs to be relatively close to as reliable. Red drives might not be as fast, but they are reliable and cheaper. Also designed to run 24/7.
And I have still lost significant amounts of data I can't replace. Never ever use NTFS compression if you want to restore files. And entropy seems to simply with enough time go beyond the errror correction ability of the drive, and so I lost 800gb of well organized v0 mp3s, as I used to help run a tracker that's not around anymore. What I wouldn't give to have that collection back. Two years of photos just ot lost and I didn't notice and had already did full backups by the time I did.
Some good data recovery software, but if the file headers are over zeroed out space it;s gone forever. This happens if you backup a corrupt MBR and scanning doesn't see it.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
51 Comments
Back to Article
willis936 - Tuesday, August 31, 2021 - link
> Unlike SSHDs, the OptiNAND drives do not store any user data at all during normal operation. Instead, the NAND is being used to store metadata from HDD operation in order to improve capacity, performance, and reliability.This is done on zfs and (in some proprietary cases) in btrfs. It's interesting to see the concept implemented at a lower level. More tools to work with.
TeXWiller - Tuesday, August 31, 2021 - link
I got the impression that this is not the file system metadata, but of the lower level related to the actual operation of the drive itself.willis936 - Wednesday, September 1, 2021 - link
Ah yeah probably. HDDs have been doing lots of bookkeeping for a long time. I wonder if this change (from what?) actually provides any improvement to HDD performance.coburn_c - Thursday, September 2, 2021 - link
The change is that the marketing team is now slapping it on the boxsleepeeg3 - Friday, September 3, 2021 - link
Exactly. How is this any different from hybrid hard drives (10+ years ago)?https://www.tomshardware.com/reviews/seagate-momen...
"Hybrid hard drives (H-HDDs) arrived in 2007, and they disappeared soon after because there was hardly any real benefit."
vFunct - Thursday, September 2, 2021 - link
It's not clear what exactly is the metadata information is being stored here.Also, rather than using actuator heads, why aren't drive heads a single linear piece solid-state hardware that has thousands of electronically addressable read/write heads on it?
coburn_c - Thursday, September 2, 2021 - link
The tracks aren't in predictable place for one, and I doubt you could pack them as tight as there are steps in the actuatorback2future - Sunday, September 5, 2021 - link
Interesting idea for reducing latency, because a rotating disk would bring magnetic information towards (mostly) each solid-state read-write-head during one rotation, if placed tangentially?Saving another moving part within hdd's?
Did not find stats for production cost of parts of hdds listed separately (also endurance and failure rate of hdd parts) or percentage of total cost of devices?
back2future - Sunday, September 5, 2021 - link
at least, the actuator would have reduced movement and acceleration latencies, if placed radially, for fine tuning between (maybe more expensive) read-write-heads?Kamen Rider Blade - Tuesday, August 31, 2021 - link
That means if 9 Platters can become 20 TBThen Toshiba's 10 Platters can become 22 TB
Since WD & Toshiba are now part of the same company, shared technology should push this forward.
Now if only they can add in 64 GiB of Optane AKA 3DXpoint as a Read/Write buffer.
Wereweeb - Tuesday, August 31, 2021 - link
WD hasn't bought Kioxia yet. And I'd rather they don't.Wereweeb - Tuesday, August 31, 2021 - link
Besides, Kioxia (Ex-Toshiba SSD) is different from the Toshiba subsidiary that manuf.s HDD'sKamen Rider Blade - Tuesday, August 31, 2021 - link
My bad, I meant HGST, not Toshiba.flyingpants265 - Tuesday, August 31, 2021 - link
One 22TB drive could hold pretty much the entire NTSC library of console videogames including Xbox 360/PS3. That will be fun for emulation. PS4/XBO+ and PC are much bigger, but still reasonable numbers.Two drives could hold 44,000 movies in 1080p. There are fewer than 44,000 movies with more than 1000 votes on IMDb.
Most popular digital media will fit inside 300TB = 14 drives. Best case $4500 in today's money. 600TB would be even better. Music, TV, comics, books, and the text internet would be good collections too.
sheh - Tuesday, August 31, 2021 - link
1GB per movie is bad quality.10GB is a more likely compromise, so 2,200 movies per drive.
But that's still not the 20-30GB per movie you have on Bluray, or x2-3 that much for 4K.
flyingpants265 - Wednesday, September 1, 2021 - link
Thanks, I think most of us know the difference between 1GB and 25GB rips.sheh - Thursday, September 2, 2021 - link
By that logic, might as well go for 100,000 movies. :)flyingpants265 - Friday, September 3, 2021 - link
Why not 1 million? There are about that many IMDb titles. In such a large database I think some loss in quality would be acceptable.yacoub35 - Friday, September 3, 2021 - link
sheh is correct, 1GB is far too low an estimate for standard length (90-120 minute) HD movies encoded without excessive compression. 10GB may be a little high but calculation estimates of this sort should always be done conservatively by providing a little extra buffer overhead, so 10GB is much better to use as a a per-media-file size estimate than 1GB.anad0commenter - Wednesday, September 1, 2021 - link
There's still plenty of Europe and Japan exclusive games however. And I have OCD and need to store all of them, so 22TB is nowhere near enough for the PS3 set alone :(flyingpants265 - Wednesday, September 1, 2021 - link
So buy two? 600TB is really small for a distributed database anyway.Kamen Rider Blade - Tuesday, August 31, 2021 - link
Don't forget to implement multi-Actuator on a single Actuator Arm and Dual Actuators in the same 3.5" body.I see enough empty volume that with some re working, you could fit in 2x Actuator Arm Stacks with multi-actuators per arm leading to huge Linear R/W performance increases.
ganeshts - Tuesday, August 31, 2021 - link
Dual actuators are already in production in some of Seagate's drives. I think all vendors will eventually adopt the scheme in order to scale sequential bandwidth / IOPS with capacity. For now, WD believes it is not cost-effective to implement dual actuators outside their R&D labs.WD does have the triple-stage actuator in their single arm that enables finer positioning (which in turn means tracks can be bunched even closer together) in order to increase areal density.
Kamen Rider Blade - Tuesday, August 31, 2021 - link
I've seen smaller HDD's with smaller base motors and housings.Given the volume in a standard 3.5" HDD, you can easily fit at least 2x sets of arms.
If you shrink the motors for the arms and the base controller stack, you can get 4x stacks of arms.
But that requires them to re-engineer the stack for volumetric efficiency.
Then stack on Multi-Actuator per Stack.
Eventually, you might be able to get it down to each Actuator arm is fully independent of the other arm and have massive parralelism.
But that requires even more engineering.
flyingpants265 - Wednesday, September 1, 2021 - link
Maybe I'm missing something, but we should be going back to 5.25" as it gives a large (50%?) surface area boost per platter, and the Google hard drive study suggested thicker drives with many more platters per drive. With this we're already looking at 40TB drives easily. Lowering rotational speed is no big deal.The_Assimilator - Friday, September 3, 2021 - link
Good luck finding a 5.25" bay on modern desktop chassis.PeachNCream - Sunday, September 5, 2021 - link
A quick Amazon search yields a fair number of cases that feature 5.25" bays. I must be very lucky. I'd better go buy lottery tickets or something given my luck is so good.ballsystemlord - Tuesday, August 31, 2021 - link
So, will the iNAND solve the SMR delayed write problems?I don't claim this article said anything about it. I'm curious, because that's the major problem with SMR and if the heads had to move less because you have the metadata in iNAND that would at least help things.
rygaroo - Tuesday, August 31, 2021 - link
The article mentions potential SMR capacity improvements, but doesn't mention performance. My guess is that it would provide a negligible boost. Once your drive fills up and your write cache fills up, every random write still requires reading several tracks + writing back several tracks. I'm not sure how many consecutive tracks are shingled in each shingled region... but let's say 5 for an example, so 10 revolutions per write at 7200rpm is ~83ms per write or 12 IOPS. Double the tracks per shingled region to get more capacity benefit from the shingling and you half the random write IOPS. A larger write cache might help you get in a few more performance writes before you hit the read-modify-write wall, but you'll still hit it eventually. SMR is pretty much not an option if you have a random write heavy workload. If any of my assumptions are flawed, please let me know.abufrejoval - Wednesday, September 1, 2021 - link
The only flaw would be to put a random write heavy workload on HDDs these days: that's become "you're using it wrong".The whole SMR debate really mostly centers around the RAID/ZFS rebuilds where you couldn't really avoid it and the upper layers (controller and FS) didn't know about SMR.
If they were SMR aware, rebuilds might perhaps be done at shingle granularity and here big flash (perhaps better outside the drive) would help.
Hardly any storage is designed to support constant maximum write pressure, because there isn't much you can save with optimizations. NAND could mostly just push the first level "SLC cache" for SMR much further so the SMR write amplification becomes less noticeable in "average scenarios".
If they use NAND to implement the fulll SMR write buffers, they better have lots of really good NAND. And I'd probably want a fail-back mode, where exhausted NAND just gets bypassed and won't result in a complete drive failure.
Squeaky'21 - Wednesday, September 1, 2021 - link
abufrejoval, you mention 2.5" 8TB CMR HDD. Yes, that would be lovely and I'd be an immediate buyer! FYI, I have just received email confirmation from WD, here in Australia, that their 4TB and 5TB Passport drives are SMR. Getting this confirmation from a senior tech support guy was like getting blood from a stone. I went through 4 stages of attempts with different levels of techs and a dozen emails back/forth. This fonal guy was rather matter of fact, clear and concise and confirmed it completely. This promptly put a stop to my planned purchase of four of the 5TB models. Just so you know!watersb - Tuesday, August 31, 2021 - link
The key role for this embedded NAND is to assist, and possibly actively tune, the basic mechanical challenge of hitting an impossibly small target on rotating platters:> repeatable run out (RRO) recording of the head jitter / error position as the spindle revolves. This data (running into multiple gigabytes) is generated in the factory during manufacturing. It is typically stored in the disk, taking up space that could have potentially been used for user data. The OptiNAND architecture moves this to the NAND in the EFD.
Discussed in the slides here, at around the third or fourth slide.
Apparently, the mechanical “map” of each platter takes gigabytes of data, particular to each hard disk, each platter and segment. Although manufacturing tolerances on these things are amazing, each physical device has its bumps… to say nothing of non-uniform (at this scale) heat explanation.
The second role for this NAND is to buffer adjacent sectors (on nearby tracks as well as the same track) after writing… I didn’t realize how we’ve hit the wall there, having to re-write entire neighborhood of sectors every 6 writes or so:
> One of the downsides to increasing areal density by increasing the TPI is the need to do more frequent refreshes. From refreshing once in 10000 write operations in early HDDs, the narrow tracks now need to be refreshed as frequently as once every 6 writes.
mode_13h - Monday, December 6, 2021 - link
> The second role for this NAND is to buffer adjacent sectorsNo, I'm pretty sure it doesn't actually buffer data to be written. I think it's probably used to keep per-sector write statistics, so the drive can allocate sectors more intelligently and knows when to schedule rewrites.
abufrejoval - Wednesday, September 1, 2021 - link
Fascinating insights into how HDDs are transforming!With rewrites of adjacent tracks dropping to 1 in 6, the real difference between CMR and SMR really starts to disappear.
The whole question as to where you'd want to put the intelligence still doesn't have a good answer: Here it's the drive, the smart RAID controller was in favor for a long time and I guess hyperscalers might just go for a (specialized?) server anyway, up to the point where the rewriting of shingles it a decision made by the host not the drive.
I'm surprised flash is judged fast enough for the operational data, but then again, the proper parts can be picked up in all those milliseconds that it take the heads to move/switch or the sectors to rotate: you can do a lot of compute in milliseconds these days!
For my personal use at these capacity points the RAID6/ZFS2 rebuild debate becomes a bit mote and a mirror set, for all its wasted capacity becomes the maximum that makes any sense. SMR would compensate that a little and I guess it's just time to complete the mental switch: HDD is tape and SSD is HDD.
I still wouldn't mind seeing 2,5" CMR drives with 8TB, but I guess that species has gone extinct.
thegreatyashu - Sunday, September 5, 2021 - link
A travesty the largest 2.5"drive is 5tb, and too thick for most laptops.thegreatyashu - Sunday, September 5, 2021 - link
I use 2.5" drives for cheap non volatile raid for booting desktops. All they really good for. WD has a single platter 1tb super thin, draws less currrent than ssd, 128mb cache. They fit anywhere in a PC case.thegreatyashu - Sunday, September 5, 2021 - link
Mint linux will respect your motherboard raid. Other distros still will seee them as only separate.mode_13h - Monday, December 6, 2021 - link
> With rewrites of adjacent tracks dropping to 1 in 6,> the real difference between CMR and SMR really starts to disappear.
Not really. If the controller allocates sectors so they're roughly grouped by the number of writes, then you're only rewriting a zone 1/6th of the time. That's still way less than *every* time.
mode_13h - Monday, December 6, 2021 - link
> HDD is tape and SSD is HDD.No, SSDs don't have the same potential for cold storage as HDDs.
Squeaky'21 - Wednesday, September 1, 2021 - link
abufrejoval, you mention 2.5" 8TB CMR HDD. Yes, that would be lovely and I'd be an immediate buyer! FYI, I have just received email confirmation from WD, here in Australia, that their 4TB and 5TB Passport drives are SMR. Getting this confirmation from a senior tech support guy was like getting blood from a stone. I went through 4 stages of attempts with different levels of techs and a dozen emails back/forth. This fonal guy was rather matter of fact, clear and concise and confirmed it completely. This promptly put a stop to my planned purchase of four of the 5TB models. Just so you know!Squeaky'21 - Wednesday, September 1, 2021 - link
I've just purchased an 18TB WD GOLD HDD and have a few 14TB Ultrastars in my home server setup. My opinion is that they are simply the very best HDDs that have ever existed on the planet. I expect I will have these same drive a decade or two into the future - they are SO well made. Let's hope the helium stays contained eh?! I do a lot of 4K video editing. I bought my 18TB GOLD due to the fact all the reviews and WD itself said that was the biggest CMR drives could ever get. The 20TB Ultrastar was, of course, SMR. So NOW they come out and say 20TB will indeed be CMR too!! I don't know about you but it all sounds a bit 'too' experimental if they are now able to cram an extra 2TB onto the 18TB CMR disk simply because they have a 64GB slither of flash on the side! If I am missing anything I would really appreciate being brought up to speed on this. Cheers all - oh, my 1st post here tonight! :-)ganeshts - Wednesday, September 1, 2021 - link
The top-capacity HDDs (irrespective of the vendor) for the last few years are quite reliable because the platforms they are based on have been created mainly for enterprise customers. Unless those pass enterprise / cloud-deployment qualification, they do not make it out to the other product lines - simply because the selling price is too high for non-power users to adopt them, and there is no incentive for the vendors to face a backlash (likely that top capacity sales are not a big chunk of the revenue from the consumer segment).Brane2 - Thursday, September 2, 2021 - link
Meh. Lipstick on a pig. Both, the product and thinly veiled advert, masked as an article.Where are substantial improvements ?
Like multiple headstacks, with multiple active heads within each stack ?
Also, capacity jump is minimal. Just 2 extra TB and those are gained by losing signal integrity, because they are pushing data density on the same process.
They are also cooking the public to gradually accept SMR through adjacent track degradation trick.
By squeezing tracks ever more closer, far more thatn what the haeds can reliably adress, they are squeezing SMR into all drives.
Spunjji - Friday, September 3, 2021 - link
"Where are substantial improvements ?Like multiple headstacks, with multiple active heads within each stack ?"
Those kind of improvements will require improvements like the ones detailed to even be feasible. Even then, they're still non-trivial, and they're going to add significantly to costs.
The_Assimilator - Friday, September 3, 2021 - link
Imagine if they spent that time and money into just making higher-capacity NAND instead of beating the dead horse that is mechanical disc technology.mode_13h - Monday, December 6, 2021 - link
Plenty of people *are* putting a lot of time, energy, & money into higher-capacity NAND. However, HDDs still offer more GB/$ and modern, high-capacity NAND cannot be used for cold storage.thegreatyashu - Sunday, September 5, 2021 - link
I agree about the WD golds. I thought 12tb was perfect price/size and these drives are phenomenal. Fast, too. Faster than red 12tb drives, or esp the white label 12tb drives they use in their externals. I am constructing a NAS, and raid 01 or 10 seems the way to go for 24gb of storage. I think raid 5 is not used as often as the past.I have tried to hold onto all my data going back to the 90s. I have been through this a few times. I have to have two copies at least other than live. Older drives come in handy, as the 2nd copy can be jbod, but first copy of the live storage needs to be relatively close to as reliable. Red drives might not be as fast, but they are reliable and cheaper. Also designed to run 24/7.
thegreatyashu - Sunday, September 5, 2021 - link
And I have still lost significant amounts of data I can't replace. Never ever use NTFS compression if you want to restore files. And entropy seems to simply with enough time go beyond the errror correction ability of the drive, and so I lost 800gb of well organized v0 mp3s, as I used to help run a tracker that's not around anymore. What I wouldn't give to have that collection back. Two years of photos just ot lost and I didn't notice and had already did full backups by the time I did.Some good data recovery software, but if the file headers are over zeroed out space it;s gone forever. This happens if you backup a corrupt MBR and scanning doesn't see it.
We need much cheaper cloud services.
thegreatyashu - Sunday, September 5, 2021 - link
I have backups of every c drive from past PCs from the last 15 years. Some boot on VMware.Anymoore - Monday, September 13, 2021 - link
What was used to store the metadata before 3D NAND? Serial Flash? DRAM?mode_13h - Monday, December 6, 2021 - link
It is/was stored on the disk platter, itself. Interleaved with the actual data, I guess.