Seagate Unveils 10TB Helium Enterprise Drive

Megalith

Megalith

24-bit/48kHz
Staff member
Joined
Aug 20, 2006
Messages
13,000
This follows the release of WD’s 10TB drive back in December. Do you think that amount of data is too much to entrust to a single HD?

The new, robust 10TB Enterprise Capacity 3.5 HDD provides maximum storage capacity for easy system integration by using the standard 3.5-inch CMR design. Incorporating seven platters and 14 heads, the drive seals in helium to create a turbulence-free, quiet environment, decreasing both friction and resistance on the platters and delivering the industry’s lowest power/TB ratio and weight specifications for a 10TB HDD. Offering 25 percent more density to help businesses dramatically increase petabytes per rack, the drive delivers higher performance and reduced power and weight.
 
Instead of making clicks when it dies, does it instead make funny squeaky dying sounds due to the helium?


Also, this probably makes me an awful person, but when they first turned on the LHC I made a joke about the funniest disaster to ever happen would be a helium leak there because of all the cries for help would sound like munchkins. At least Helium has a sense of humor unlike Nitrogen. It just takes ya out with no warning and nothing funny.
 
The cries for help would be the amount of porn lost from a failed hard drive.
 
"Do you think that amount of data is too much to entrust to a single HD?" Yes if Seagate is the manufacturer....Although the 340MB ST3385A I got as a refurb from Midwest Micro in 1993 is still running fine in my DOS rig for old games.
 
inb4 "hurr durr that's a lot of data to lose"

Edit: too late
 
wonder what the life span is.... He leaks out of EVERYTHING even metal
 
I'm worried about how we keep finding new ways to burn through helium... as if people confused helium and hydrogen.

Charge $100 for a helium balloon and things will calm down!
 
Synful Serenity said:
"Do you think that amount of data is too much to entrust to a single HD?" Yes if Seagate is the manufacturer.... Although the 340MB ST3385A I got as a refurb from Midwest Micro in 1993 is still running fine in my DOS rig for old games.
Click to expand...
Was my first thought too
 
andrewleaster said:
I'm worried about how we keep finding new ways to burn through helium... as if people confused helium and hydrogen.

Charge $100 for a helium balloon and things will calm down!
Click to expand...
Helium is the second most abundant element in the entire universe.

Helium is just getting more expensive because of supply and demand. Its not that we would run out of helium any time soon, but that we usually don't ever drill (where its easiest to get) for helium specifically. We drill for natural gas and what not, and it usually has a lot of helium in it too, but we only extract so much every year, and while it used to be that there was little use for the gas so it was cheaper than crap, we are finding lots of uses for it now. So as demand raises compared to the regular annual supply, the once virtually costless commodity goes up a bit in price.

But that doesn't mean that in 100 years we'll run low on helium or anything.

The only likely problem would be the helium prices would rise enough that we'd stop wasting it on balloons for parties.
 
I wouldn't buy a seagate drive, but the 10Tb WD would be sweet. 5 in a raid 5 gives you 40Tb. Yum. I got a 16Tb raid 5 atm so i should be good for awhile. maybe in 3 or 4 years these will be cheaper and more reliable and I will spring for some. These or SSD's of the same capacity. If SSD's are still capacity limited I will stick with mechanical.
 
I'd like to see a drive with built in raid. say each platter is 1TB, 4 platters in a raid setup internal to the drive. The PC sees it as one drive.
 
Would like to see the reliability of these :p. So awesome more and more 10TB drives are coming out now though.
 
The good: 10 TB HDD
The bad: it's a Seagate (can be good or bad)
The ugly: what happens when it fails (10 TB is still a lot of data)
 
Nice to see a 10TB drive, but from Seagate? No way would I trust it.
I also don't think I'd trust any drive with 7 platters/14 heads. Too much to go wrong.

The WD 6TB drives have 5 platters, and that's about as high as I trust.

I'm still not going over 4TB drives for my production servers (Raid 6 or 10).
I've just started using the 6TB drives in my backup server because I needed the space, and it was a lot cheaper than adding an external drive bay.
The backup server isn't as critical since everything is copied to tape, so even if the entire raid failed, I can rebuild the TB's of data in less than a day.
 
amdfan25 said:
The good: 10 TB HDD
The bad: it's a Seagate (can be good or bad)
The ugly: what happens when it fails (10 TB is still a lot of data)
Click to expand...

Given that it's an enterprise drive, I imagine you're meant to buy 2+ and set up a redundant RAID system.
 
Galvin said:
I'd like to see a drive with built in raid. say each platter is 1TB, 4 platters in a raid setup internal to the drive. The PC sees it as one drive.
Click to expand...

That's exactly how they work. Data is striped down the heads, then around the track, then across the cylinders. It doesn't fill up one platter then move to the next one.


What I've always wondered was why they don't put multiple independent read head assemblies in there. a drive could service multiple read requests simultaneously at the platter instead of just in the queue. (I know there's not much room in current form factors... but there's no reason they couldn't change that).
 
MavericK said:
Given that it's an enterprise drive, I imagine you're meant to buy 2+ and set up a redundant RAID system.
Click to expand...

Not in raid 1, Nope nope nope. You will get a failure after you replace the drive during a rebuild.

Been there done that 5 times at work. Thankfully we have back ups but still.
 
Not in raid 1, Nope nope nope. You will get a failure after you replace the drive during a rebuild.
Click to expand...

This is why you do weekly scrubs with all raid arrays.
 
GoodBoy said:
I wouldn't buy a seagate drive, but the 10Tb WD would be sweet. 5 in a raid 5 gives you 40Tb. Yum. I got a 16Tb raid 5 atm so i should be good for awhile. maybe in 3 or 4 years these will be cheaper and more reliable and I will spring for some. These or SSD's of the same capacity. If SSD's are still capacity limited I will stick with mechanical.
Click to expand...

If you are running drives that big in RAID5 then you are the last person on the planet with finger wagging rights.
 
With those Dell T20 servers being so rediculously dirt cheap, I just decided that my best performance/redundancy was to just use both as HTPCs connected to two different TV's, and have them mirror each other with software backups where the slave always tries to look like the master (adding and removing files on each sync). Granted it only has four 3.5" bays, but these days with 10TB storage space... who cares? And since I always keep my electronics behind a UPS at each station, the chances of having a simultaneous failure of both systems is virtually none. Heck, drives seem so reliable these days that I haven't had a drive failure in AGES anyway, so in the off chance I finally have one die on me, I'll just 2-day Amazon a replacement and rebuild it from its clone.
 
Elios said:
multiple off site backups and off line backups
Click to expand...

The speed of these drives doesnt change much with size, only if the density improves, which it doesnt if more platters are used instead.
It takes a hell of a long time to repopulate 10TB.
The longer a drive is servicing requests, the larger the chance of a failure.

ryan_975 said:
That's exactly how they work. Data is striped down the heads, then around the track, then across the cylinders. It doesn't fill up one platter then move to the next one.


What I've always wondered was why they don't put multiple independent read head assemblies in there. a drive could service multiple read requests simultaneously at the platter instead of just in the queue. (I know there's not much room in current form factors... but there's no reason they couldn't change that).
Click to expand...
By RAID it was fairly clear he meant from the performance perspective as you addressed in the second part of your post.

Its my feeling too that internal RAID of platters should be handled on the drive.
At least on high end drives.
They would compete with SSDs to some extent and reduce the gravity of the switchover, saving their businesses!
 
Nenu said:
The speed of these drives doesnt change much with size, only if the density improves, which it doesnt if more platters are used instead.
It takes a hell of a long time to repopulate 10TB.
The longer a drive is servicing requests, the larger the chance of a failure.


By RAID it was fairly clear he meant from the performance perspective as you addressed in the second part of your post.

Its my feeling too that internal RAID of platters should be handled on the drive.
At least on high end drives.
They would compete with SSDs to some extent and reduce the gravity of the switchover, saving their businesses!
Click to expand...


It was fairly clear he was talking about writing to or reading from all platters at once like RAID does with each drive in an array. They already do that.
 
ryan_975 said:
It was fairly clear he was talking about writing to or reading from all platters at once like RAID does with each drive in an array. They already do that.
Click to expand...

No they dont.
Otherwise performance would increase with number of platters.
 
ryan_975 said:
That's exactly how they work. Data is striped down the heads, then around the track, then across the cylinders. It doesn't fill up one platter then move to the next one.


What I've always wondered was why they don't put multiple independent read head assemblies in there. a drive could service multiple read requests simultaneously at the platter instead of just in the queue. (I know there's not much room in current form factors... but there's no reason they couldn't change that).
Click to expand...

Something like this?

(photoshop obviously)

p00020001_dualHeadHDD_web.jpg
 
Nenu said:
No they dont.
Otherwise performance would increase with number of platters.
Click to expand...

Okay, maybe "at the same time" isn't accurate. but it does spread sector operations across the platters before it moves to another area. RAID doesn't even write to disks at the same time. It send 4KB-128KB+ to one disk, the next 4KB-128KB+ to the next drive and so on.
 
ryan_975 said:
Okay, maybe "at the same time" isn't accurate. but it does spread sector operations across the platters before it moves to another area. RAID doesn't even write to disks at the same time. It send 4KB-128KB+ to one disk, the next 4KB-128KB+ to the next drive and so on.
Click to expand...

RAID increases performance by performing simultaneous writes.
If it didnt do that, it couldnt increase performance.
 
Nenu said:
RAID increases performance by performing simultaneous writes.
If it didnt do that, it couldnt increase performance.
Click to expand...

Yes it would. a controller spends most of it's time waiting for a drive's physical activities to complete. With RAID, instead of sitting there idle while a disk is busy, it can move to another disk in the array and send it some data, then the next and so on. This gives the illusion that you're not having to wait as long for data to be written making it effectively faster.
 
ryan_975 said:
Okay, maybe "at the same time" isn't accurate. but it does spread sector operations across the platters before it moves to another area. RAID doesn't even write to disks at the same time. It send 4KB-128KB+ to one disk, the next 4KB-128KB+ to the next drive and so on.
Click to expand...

It technically could work the way you first described, the issue being that you'd have to get data that was lined up underneath each read / write head on the exact same sector on each platter. I don't know enough about the details but I could see in certain situations that might actually occur on large file transfers that were single queue depth. Then again maybe the drive couldn't actually handle all of the heads sending data back at the same time, I'm not really sure.

The main issue is that even if you are a sector off then the drive is going to have to wait until the platter comes around again to move over in order to read that data. In theory if you could split the head in half you could potentially decrease seek times on higher queue depths if one head could seek out data on the bottom half of the platters while another did so on the top. I wouldn't be surprised if someone has already tried that though and determined it wasn't cost effective or gave a bit enough boost to warrant doing so. Most drives under 1TB these days are likely single platter designs, so there is only a pair of heads. It's only these really large drives where there might be 10 heads reading data where it might make sense.
 
ryan_975 said:
Yes it would. a controller spends most of it's time waiting for a drive's physical activities to complete. With RAID, instead of sitting there idle while a disk is busy, it can move to another disk in the array and send it some data, then the next and so on. This gives the illusion that you're not having to wait as long for data to be written making it effectively faster.
Click to expand...
Read up on RAID performance, you have some misconceptions.

The controller operates orders of magnitude faster than the drives creating instructions and sending them to the drives.
The controller can send data to a string of drives before the drive head has even finished its first movement on the first drive.
All the drives operate simultaneously in every aspect. The may start operations nS apart but that has very little impact.

The fact a drive takes time to perform its operations is expected as it cannot perform 2 operations at once. This is elementary.
And the time it takes to issue the next instruction is nS is queued in the drive anyway while it is still completing a task.
This allows the drive to operate with its maximum throughput, not matter how many drives are in the array.
It always has a full queue of instructions, there is no waiting.
 
Mathematician said:
Its funny to me that people are referring to 10tb drives as 'very large' once they get used to it and much larger onres are released they won't even remember thinking they were large. ��
Click to expand...

I remember a time when I couldn't imagine filling up a 1.2GB drive. :D
 
Mathematician said:
Its funny to me that people are referring to 10tb drives as 'very large' once they get used to it and much larger onres are released they won't even remember thinking they were large. ��
Click to expand...

The magnitude of data is one thing, its a shit load to lose if you dont do a backup.
Its also a huge time investment to repopulate a drive of that size.
Even if it could sustain 200MB/s, it will take over 13hrs to fill !!

For home users thats a royal pita.
For business users without dual fail over, thats a substantial time until your backup is ready.
 
I don't entrust ANY data to a single HD.

At the very least there needs to be some redundancy.

Now, getting multiples of these and setting up a raid setup with redundancy might not be bad.
 
Megalith said:
Do you think that amount of data is too much to entrust to a single HD?
Click to expand...

Who puts 10tb of data they don't want to lose on 1 drive? :D

Do we have to do this every time an increased capacity hdd comes out? Moron
 
LOCO LAPTOP said:
Not in raid 1, Nope nope nope. You will get a failure after you replace the drive during a rebuild.

Been there done that 5 times at work. Thankfully we have back ups but still.
Click to expand...

I've never had a Raid fail during the rebuild other than a problem with the replacement drive. If the replacement is bad, I replace it again and the rebuild starts over.

I do have the raid controllers set to check the raid every week, which is probable why I haven't had a problem with unreadable sectors of the drives.
However, I've now switched to Raid 6 on my larger raids (over 10TB) as there is too much risk in a rebuild with Raid 5.
 
Galvin said:
I'd like to see a drive with built in raid. say each platter is 1TB, 4 platters in a raid setup internal to the drive. The PC sees it as one drive.
Click to expand...

How about no. RAID5 is a performance and uptime multiplier, not a reliability multiplier.
 
DPI said:
How about no. RAID5 is a performance and uptime multiplier, not a reliability multiplier.
Click to expand...

Yeah, I don't see how this would work. It seems unlikely to me that one platter would fail and the others keep working. After all the arm for the heads is shared...

That, and what would you do if one platter failed? It's not as if you can replace the failed platter...
 
You must log in or register to reply here.
Back
Top