WD has some new drives. How are they cramming more data per platter diks in there? It's not clear if they use the WHAMR or something.
Last edited:
WD 18TB and 20 TB
- Thread starter LunarMist
- Start date
Chewy509
Wotty wot wot.
?? The actual NAME of the product contains the recording technology ??
18TB CMR - Conventional Magnetic Recording
20TB SMR - Shingled Magnetic Recording
?? Also super clear in the linked datasheets on the technology being used ??
"Western Digital is combining a new nine-disk platform and energy-assisted recording."
Of course we have already discussed the CMR vs. SMR recording technology, but what is the energy assist technology? I doubt it is like a Chevy.
Of course we have already discussed the CMR vs. SMR recording technology, but what is the energy assist technology? I doubt it is like a Chevy.
sdbardwick
Storage is cool
You just listed it in your previous post. HAMR, MAMR are types of energy assisted recording. Heat (energy) Assisted Magnetic Recording; Microwave (energy) Assisted Magnetic Recording.
jtr1962
Storage? I am Storage!
My guess is all the tricks to increase density cost so much to implement that we end up with the same cost per TB. All most end users care about is cost per TB.
Besides that, I don't think the volume is there any more to drive significant R&D of HDDs. SSDs in sizes most people want (i.e. 2TB or less) can be had for under $100. Very few people want or need even 10TB drives, never mind 20TB. It's basically data centers, and a few outliers like you, who need drives this big. Back in the days when it seemed HDD sizes were doubling every year, every single PC had an HDD, some more than one.
I don't think the cost per TB of the drive is the issue, but the cost of storage including the entire infrastructure.
For a home user like me the cost of the NAS is important, because over 8 bays is significantly more expensive and the size, weight, and power increase. Over 12 bays and you are looking at rack mounts that are even more costly and generally have noisy fans for home use.
In a datacenter the energy and footprint are important considerations.
For a home user like me the cost of the NAS is important, because over 8 bays is significantly more expensive and the size, weight, and power increase. Over 12 bays and you are looking at rack mounts that are even more costly and generally have noisy fans for home use.
In a datacenter the energy and footprint are important considerations.
Last edited:
I read that the Toshiba 20TB drive has use of the MAMRs, the FC-MAMR species. Maybe it uses the Flux capacitor. 
32TB HAMR drives on the horizon with 40 and 50TB in the pipeline
www.tomshardware.com
Seagate Reveals HAMR HDD Roadmap: 32TB First, 40TB Follows
Seagate expects its 24TB HDD to be its last PMR product.
www.tomshardware.com
Mercutio
Fatwah on Western Digital
WD says it'll have a 28TB drive without any of that weirdo HAMR stuff by next year. Not as exciting as a 32TB drive with laser beams in it, but CMR is decades-proven technology and that probably has some appeal as well, even though I weep for the poor bastard who gets stuck with WD's version of SMR on an almost 30TB volume.
Seagate has an Exos X24 listed (24TB). Is that not a natural drive?
Seagate Exos X24 Hard Drive | Seagate US
Maximize your enterprise storage with Seagate's Exos X24 Hard Drive, featuring a massive 24TB capacity and superior rack-space efficiency for reliable performance.
www.seagate.com
Mercutio
Fatwah on Western Digital
There are also 30TB+ SSDs if you have any spare kidneys sitting around.
We were on travel during the WD announcement. I was thinking of when it would be a good time to replace smaller hard drives. 24TB is interesting, but just a small increment. Did you see them in the wilds yet?Seagate has an Exos X24 listed (24TB). Is that not a natural drive?
Seagate Exos X24 Hard Drive | Seagate US
Maximize your enterprise storage with Seagate's Exos X24 Hard Drive, featuring a massive 24TB capacity and superior rack-space efficiency for reliable performance.www.seagate.com
Mercutio
Fatwah on Western Digital
Actually, there's a standard for even higher density SSDs, called 3.S or 3.L, for when you really want a single 100TB+ drive right now.
I'm sure Linus Tech Tips will be along in about a month making a video about how they built a new 32PB storage server with 64 of them that some memory manufacturer gave them for nothing but goodwill.
jtr1962
Storage? I am Storage!
Well, SSDs are getting a lot cheaper. We're at around $30/TB for the commodity stuff. Commodity HDDs are maybe only half that at this point, but SSDs are falling in price faster than HDDs. Probably only a matter of a year or two before SSDs get cheaper than HDDs. The highest capacity stuff though seems to cost way more per TB.
HDDs have reached the same point now as LEDs. We're only seeing incremental increases in capacity, just as we're only seeing incremental increases in LED efficiency. Both things are close to plateauing out as they near theoretical maximums.
Something like a 30.72TB Micron 6500 Ionic SSD would be good for me.
However, I still need the HDDs for NAS. It would be great if an 8-bay could hold everything.
I haven't seen them yet but assumed they should hit the channels soon-ish. Agreed it's only a slight bump in capacity for HDDs but at least it's something. Maybe the 20's and 22's will come down in price.
The cheap junk is QLC with no RAM and low erase cycles and it is not $60 for 2TB that you would want to use. SSDs are not a substitute for huge HDD arrays in the Amazon or the Google either. I recall when many were saying that HDDs would be dead by 2015.
jtr1962
Storage? I am Storage!
They are practically dead for anything but data centers, and a handful of home users with huge storage needs.I recall when many were saying that HDDs would be dead by 2015.
Death Of The Hard Disk Could Come Sooner Than We Thought
New research shows SSDs are almost twice as reliable as hard disks, potentially making them more cost-effective
www.forbes.com
However, with further evidence of SSDs showing greater long-term reliability than hard disks, the economics of switching mass storage to SSDs are becoming ever more attractive, despite the upfront cost of the hardware, according to Backblaze.
The latest report from Backblaze compares thousands of hard disk and SSDs used by the company at the same stage in their lifecycles. The research found that hard disk drives had an average failure rate of 1.54%, whereas the SSDs had a failure rate of only 0.79%, making them almost twice as reliable.
The latest report is here. The older 10 and 8TB drives are not doing very well.
www.backblaze.com
The strange thing is that they are still using nearly 12,000 old 4TB drives. Wouldn't the space and power requirements compared to modern drives make that a bad idea now? It's nice that the Blazeblackers are giving out data, but is this normal usage in IT? I just can't help wondering if the Google and Amazon do that.
Backblaze Drive Stats for Q2 2023
Read the full Q2 2023 Drive Stats report is here, and we've added some data points and new drives to the data.
www.backblaze.com
The strange thing is that they are still using nearly 12,000 old 4TB drives. Wouldn't the space and power requirements compared to modern drives make that a bad idea now? It's nice that the Blazeblackers are giving out data, but is this normal usage in IT? I just can't help wondering if the Google and Amazon do that.
jtr1962
Storage? I am Storage!
Maybe the 4TB are more reliable? Yes, it seems strange they didn't upgrade to larger drives. They must have run the numbers and found some reason to stick with the older drives.The latest report is here. The older 10 and 8TB drives are not doing very well.
Backblaze Drive Stats for Q2 2023
Read the full Q2 2023 Drive Stats report is here, and we've added some data points and new drives to the data.
The strange thing is that they are still using nearly 12,000 old 4TB drives. Wouldn't the space and power requirements compared to modern drives make that a bad idea now? It's nice that the Blazeblackers are giving out data, but is this normal usage in IT? I just can't help wondering if the Google and Amazon do that.
Those HGST MegaScale drives were cheap, designed for light loads, had only 800kh MTBF, 10^-14 UBER, and only a 3-year warranty. I suppose the low power from the 5400 or 5700 RPM produced a lot less heat than the typical "Hibachi" drives of the era, but even at 5-6W for 4TB that's a lot compared to 6-7W for a 20TB DC HC560 now.
The WD 24TB CMR are shipping.
CougTek
Hairy Aussie
24TB Seagates are available now, such as the EXOS and Skyhawk AI. I'm not sure what an AI drive is. Of course there are some FW differences for video, but no indication that the drive does any AI computations. The WD Ultrastar DC HC580 is not really our yet.
I agree, the AI part in the name seems more like a marketing choice than actual AI abilities from within the drive. The drives appear to be slightly more durable with longer MTBF and TB written and offers more caching and a tiny bump in transfer rates compared to the non-AI version (1.9% faster).
Chewy509
Wotty wot wot.
I would assume the 'AI' part means the firmware is designed for typical IO workloads found when used with AI software. Similar to how drives for video systems is designed more for linear IO, vs random IO. And HHDs for NAS may be tweaked for more random IO and quieter running, and so on.
But the video drive model (skyhAWK) is the one optimized for AI.
Here is some marketing stuff about the AI. It's not clear if the ImagePerfect AI is a drive function or software.
Rebuilding an array 3x faster seems unlikely.
Seagate Blog | Seagate US
Read the latest stories and news about hyperscale storage technology, cloud infrastructure, and beyond.
blog.seagate.com
Rebuilding an array 3x faster seems unlikely.
Chewy509
Wotty wot wot.
Based on my reading, it's firmware tweaked for AI workloads that process video... My guess it's the same firmware used for video surveillance systems (basically linear writes at low IO queue depth), but tweaked to be more responsive to random IO in regions the were written recently... (The cache holds recently written data, as opposed to a cache design to primarily focus on recently read data).
The additional MTBF numbers are most likely due to improved manufacturing rather than firmware changes, mind you I assume MTBFs numbers for a drive that performs mostly linear writes, will be different to a drive in a RAID array that does near pure random IO. (due to wear on things like the drive arm motor and actuation point, and so on).
The additional MTBF numbers are most likely due to improved manufacturing rather than firmware changes, mind you I assume MTBFs numbers for a drive that performs mostly linear writes, will be different to a drive in a RAID array that does near pure random IO. (due to wear on things like the drive arm motor and actuation point, and so on).