News New HAMR lasers could usher in 30TB+ HDDs, Seagate and Sony team up for production: Report

[Admin]

Seagate reportedly second sources heating lasers for HAMR heads from Sony.

New HAMR lasers could usher in 30TB+ HDDs, Seagate and Sony team up for production: Report : Read more

 

[hotaru251]

I am curious as to the maximum size of HDD before they are pointless due to how large size is that you can't reliably copy files within the "safe" time window in case of a drive failure to rebuild your array.

 

[jpeckinp]

Ok maybe I missed something, but don't we already have 30TB HDD's?

 

[hotaru251]

but don't we already have 30TB HDD's?

28 i believe was highest HDD prior.

we have much larger SSD (liek 60TB i believe?) but not hdd.

 

[Vanderlindemedia]

I am curious as to the maximum size of HDD before they are pointless due to how large size is that you can't reliably copy files within the "safe" time window in case of a drive failure to rebuild your array.

Would take a dual actuator at some point to "cope up with" the ever demanding speeds and data intensity.

At some point NVME's will make HDD's obsolete. The only selling point a HDD has is the space.

 

[Grobe]

Would take a dual actuator at some point to "cope up with" the ever demanding speeds and data intensity.

That is a valid design challenge for the manufacturers. However, I'm more interesting in how to mitigate the SMR bad random write performance. The only kind of solution I'm aware of today is there is a specific way to format ext4 file system.

On that note, the new hamr based disk drives - wonder if those disks are CMR based? I couldn't tell from the article.

 

[Vanderlindemedia]

Or use a SSD/HDD hybrid design, where the SSD part operates as a cache level in between.

 

[leclod]

I am curious as to the maximum size of HDD before they are pointless due to how large size is that you can't reliably copy files within the "safe" time window in case of a drive failure to rebuild your array.

I understand your question. But being rational, how can this be an issue when HDDs are made to run for years. What's 1 day/2days against 5+years ? How is 2days much worse than 1day ? (or say 6days against 3days). Anyway those data centers can do the math.

 

[leclod]

That is a valid design challenge for the manufacturers. However, I'm more interesting in how to mitigate the SMR bad random write performance. The only kind of solution I'm aware of today is there is a specific way to format ext4 file system.
On that note, the new hamr based disk drives - wonder if those disks are CMR based? I couldn't tell from the article.

I'm pretty sure CMR stays the standard while SMR can be used for niche products. Their are even hybrid drives where you can choose.
Nobody (nearly) wants a bigger drive with much slower transfer rates.

 

[minnion]

28 i believe was highest HDD prior.

we have much larger SSD (liek 60TB i believe?) but not hdd.

100tb SSD, but they are around $40k per drive.

 

[hotaru251]

How is 2days much worse than 1day ? (or say 6days against 3days). Anyway those data centers can do the math.

because the longer it takes the more chance of data loss.
if you take days to rebuild an array and lose more drives that is a big issue.

 

[leclod]

because the longer it takes the more chance of data loss.
if you take days to rebuild an array and lose more drives that is a big issue.

I know that. I meant ...you know what I meant.

 

[ginandbacon]

Ok maybe I missed something, but don't we already have 30TB HDD's?

The biggest hard disk drive right now and don't want to wait for the aforementioned 30TB drive to release, is the Western Digital has a 24TB hard drive for around $650. The WD Gold Enterprise class SATA HDD is a data center HDD but will be more adequate for most usage.

I expect this to be between 700 and 800 dollars as WD is releasing either 26 or 28GB drives sometime this year.

The 100TB Exadrive from Nimbus has been out for almost 4 years. To my knowledge, nobody has been able to top that. It's in a 3.5" form factor and costs around 40K now. Linus got his hands on one a few years ago so he's got a video on YouTube. It's going to be a while before 20 to 30TB drives are out at a reasonable price as these are enterprise drives to actually save money on licensing for enterprise software the most part

View: https://youtu.be/ZFLiKClKKhs?si=hGmpy1_-w6hiY8Xl

 

[ginandbacon]

I am curious as to the maximum size of HDD before they are pointless due to how large size is that you can't reliably copy files within the "safe" time window in case of a drive failure to rebuild your array.

I question if this will ever be an issue. Especially with solid state drives. Also, I am pretty sure any decent NAS (especially enterprise level) would continue the transfer, as no data is lost. Especially depending on what RAID you are running where you can lose 2 or more drives and still have zero data loss.

A 1Gbps switch can transfer a gigabyte in roughly ten seconds, so six a minute, 360 an hour, or 8.6TB in 24 hours. How many single files do you have that large? Anyone who did would be running a 10Gbps network also. If the data transfer stopped half way through, half the files would still be on the new server. The only time lost would be the time between it stopping and the user noticing it stopped and restarting the copy of half the data that didn't make it. Lastly, hot spare drive bays exist.

A hot spare is a standby drive. When a drive crashes and results in storage pool degradation, the hot spare drive can automatically take the place of the defective drive. This allows the storage pool to repair itself and return to a healthy status in a shorter amount of time.