Sign in with
Sign up | Sign in
Your question
Closed

Exclusive: The Dual Actuator/Head Harddrive

Last response: in News comments
Share
July 17, 2009 8:03:41 PM

But really, who wants a 15,000rpm velociraptor to be turbo charged to a 'simulated' 30,000rpm? Format was already changed making the velociraptor, they could easily cram in another arm without making the drive longer than standard
Score
-11
July 17, 2009 8:05:17 PM

Because today instead of a Dual Actuator/Head setup, we use two HDDs in a RAID setup. :) 
Score
10
Related resources
July 17, 2009 8:08:44 PM

Cool Article, Keep it up Toms!
Score
18
July 17, 2009 8:11:06 PM

So we just have to ask and the answer will be revealed?!?

"Why do hotdog buns come in packs of 8 when hotdogs come in packs of 12?"

(but really, nice article) :) 
Score
17
July 17, 2009 8:23:25 PM

That first Seagate implementation isn't even what comes to mind. What I mean, is if you have one read head that does half the platter, and another for the other half then you can only increase the transfer speed if your data is completely randomly scattered. This would not help sequential reads whatsoever.

Now who would want to run a 5.25" HDD? Maybe stick two sets of platters from 2.5" drives in there and see what happens. Technically its dual actuator... Implement a RAID controller onboard and see what happens.
Score
5
July 17, 2009 8:34:22 PM

nekatrevenSo we just have to ask and the answer will be revealed?!?"Why do hotdog buns come in packs of 8 when hotdogs come in packs of 12?"



To force OCD people to buy 2 packs of hot dogs and 3 packs of buns whenever they want a hotdog, duh!
Score
13
July 17, 2009 8:51:13 PM

Just eat your hotdogs on regular sliced bread, problem solved.
Score
4
July 17, 2009 8:58:56 PM

nekatreven"Why do hotdog buns come in packs of 8 when hotdogs come in packs of 12?"(but really, nice article)


Not to derail the thread further, but the hotdog buns come in packs of 6 or 12 where i come from (QC, Canada).

Nice article.
Score
2
July 17, 2009 9:30:49 PM

Make up your mind!

Quote:
Last week we reported about Seagate's latest Cheetah 15K.7, which is the company's latest enterprise level HDD, boasting a blistering spindle speed of 15,000 RPM and a decent capacity of 600 GB.


The spec's say 15.700rpm and as any one can see so way the model indication why do you guys at toms keep calling it a 15,000rpm drive ?
Score
-9
Anonymous
July 17, 2009 9:51:22 PM

I have another question. Why don't all heads read in parallel?. That would not reduce rotational delay but it could increase throughput.
Score
-2
July 17, 2009 10:17:01 PM

hemelskonijnMake up your mind! The spec's say 15.700rpm and as any one can see so way the model indication why do you guys at toms keep calling it a 15,000rpm drive ?


Because Seagate says it's 15,000 RPM?

http://www.seagate.com/docs/pdf/marketing/po_cheetah_15...

And the ".7" in 15K.7 doesn't represent "15,700" RPM.

/ Tuan
Score
9
July 17, 2009 10:41:56 PM

Good article, great images. Well done Tuan.
Score
8
July 17, 2009 11:06:01 PM

I was wondering this too. Why can't they then use laptop platters in desktop HHD's with dual arms? might work. I think they should continue to look into this. I'm sure there are some people that would pay more for this. Oh wel, I just need to wait till SSD's get cheap now :( 
Score
4
July 17, 2009 11:09:09 PM

this seems rather stupid honestly to me by the hard drive companys from what i understand a normal 7200rpm harddrive does somewhere around 75mb/s while the SATA 3.0 goes MUCH faster even if it cost $100 instead of $50 for a 500GB harddrive there are a lot of people that will pay that for 2x the speed and there are a lot that do by making raids i am right now planing to make a 4 drive raid in my next computer i have no need of more then even 500GB of total Harddrive space yet i am going to end up with 2TB of room instead buying 2 drives that were 2x as fast in a raid would of been somthing i would do with out a problem at all.
Score
0
July 17, 2009 11:21:35 PM

I stand corrected and thanks for the bit of extra information.

./me bows down
Score
1
July 17, 2009 11:37:42 PM

It would also raise failure rates......
Score
7
July 17, 2009 11:42:22 PM

a good rule to follow for when you get the idea "Hey a really good simple way to improve XX would be _______": somebody has already thought of it before and there is a good reason why it doesnt exist. usually the answer comes from following the money
Score
1
July 17, 2009 11:56:27 PM

Great article, Tuan! This is the stuff I read Toms for :)  and of course, the usual reviews and interviews... the interesting stuff on toms :D 
Score
2
July 17, 2009 11:58:31 PM

Thanks for the feedback guys!
More articles like these coming. :) 
Hope everyone enjoys the weekend.

/ Tuan
Score
1
Anonymous
July 18, 2009 12:16:30 AM

I would say one dual head drive (eg: 500GB) would cost considerably less than 2 drives (2x250GB). It will run cooler,consume less, and will be about as fast as a raid setup of 2x250GB.

It will also be smaller in size,and could be made size compatible with normal 3,5" drives easily.

It would make sense for power efficient servers.
I can't speak of heat issues, but if you would program the driver of the disk to not have the 2 arms read/write on the same ring or sector, that may reduce performance little, but it might increase data accuracy.
Then again, if the ring has been preset, and would be a sector of about 500MB, the drive could be equipped with 256MB ram memory, that could buffer the data of the second arm (in case a linear file-read is preferred).

Anyways, as I hear it from seagate, it's a technology that never made sense back in the days, but today it does,and is easy to create.
I think it's perhaps a meager excuse why noone actually took time to further develop that theory or patent.
Instead, they buy a patent and not use it? What good does that do?
Score
4
Anonymous
July 18, 2009 12:20:15 AM

joex444That first Seagate implementation isn't even what comes to mind. What I mean, is if you have one read head that does half the platter, and another for the other half then you can only increase the transfer speed if your data is completely randomly scattered. This would not help sequential reads whatsoever.


If there's ram to buffer it could increase speeds as the first arm could read the first eg:500Mb of a movie, while the second arm continues filling a RAM space from 500 to 1024MB. This would give a slightly slower start, but end up in speeds about 2x current speeds.
Score
-1
July 18, 2009 12:49:06 AM

The dual head drive would actually be faster than 2 raid'd drives in terms of seek time, which is what actually matters in enterprise storage, raw throughput wouldnt obviously be as fast, but since its bigger anyhow, why not instead work on the materials necessary to spin the ultra-high-density platters up to 15k instead of 7200rpm? --> this would simultaneously increase throughput and seek times. (yes throughput is directly related to the disk density)
Score
0
July 18, 2009 2:08:55 AM

ProDigit80I would say one dual head drive (eg: 500GB) would cost considerably less than 2 drives (2x250GB). It will run cooler,consume less, and will be about as fast as a raid setup of 2x250GB.It will also be smaller in size,and could be made size compatible with normal 3,5" drives easily.It would make sense for power efficient servers.I can't speak of heat issues, but if you would program the driver of the disk to not have the 2 arms read/write on the same ring or sector, that may reduce performance little, but it might increase data accuracy.Then again, if the ring has been preset, and would be a sector of about 500MB, the drive could be equipped with 256MB ram memory, that could buffer the data of the second arm (in case a linear file-read is preferred). Anyways, as I hear it from seagate, it's a technology that never made sense back in the days, but today it does,and is easy to create.I think it's perhaps a meager excuse why noone actually took time to further develop that theory or patent.Instead, they buy a patent and not use it? What good does that do?


Yeah, it may have been difficult in 1996, but this is a cop out with today's tech ! Tuan, please call them on this BS and ask them to give us a reason that is concurrent with 2009. Technologically this should be easy today.
Score
-1
July 18, 2009 3:08:45 AM

royalcrownYeah, it may have been difficult in 1996, but this is a cop out with today's tech ! Tuan, please call them on this BS and ask them to give us a reason that is concurrent with 2009. Technologically this should be easy today.


Interesting point guys.

I'll talk to Seagate and see what they say about implenting such a technology using what's available and possible today.

/ Tuan
Score
0
July 18, 2009 6:31:32 AM

http://www.seagate.com/ww/v/index.jsp?name=Barracuda_2H...

We actually ordered some of these where I worked in the early 90s.
After a few months, all of them had failed (and the same problem occurred at several other places using these drive). Seagate never managed to get the mechanical problems sorted out and the drive was cancelled.

All the units we had ordered was eventually replaced by more standard drives that actually lasted more than a few months....
Score
1
July 18, 2009 7:15:41 AM

Excellent article. I always thought it would be really cool to have a 'multi-threaded' harddrive, but not for the reasons listed above.

I think it would be nice to be able to access two things on the hardrive in a truly simultaneous manner, e.g. viewing photos while copying a large file, without the random access interrupting the sequential access.

The problem is that, to really do this properly, you would need dual connections between the motherboard and the dual-head drive with both heads accessing the entire drive independently. This would require modifications to the motherboard, BIOS, operating system, etc.

Lastly, I’m not surprised by TNM’s comments above. The extra parts would drastically reduce the Mean Time to Failure of the drives.
Score
3
July 18, 2009 2:43:11 PM

I think that by the time this technology could be redeveloped to anything useful, SSD drives would have increased their speed and capacity and lowered their $/GB to levels so low that this technology would never be able to compete.
Score
1
July 18, 2009 6:49:13 PM

blackened144I think that by the time this technology could be redeveloped to anything useful, SSD drives would have increased their speed and capacity and lowered their $/GB to levels so low that this technology would never be able to compete.


I agree.

I still think seagate dosn't have a good answer to this. From a engineering standpoint it is very doable. The technical drawings of the original patent show 2 actuators/heads reading from the same platter. This is not needed in a current HDD, with 500GB per platter technology availible you could, in theory, have a 2 platter design capable of 1TB total storage and have a independent actuator/head for each platter. Like others have said this would be similar to RAID 0 in performance. As far as price goes, it would not add near as much cost as a SSD of similar capacity but reliability would be my main concern.

As you said blackened, at this time, development of somthing like this would be very time consuming and by the time a prototype was ready SSD's will be a much more viable solution.
Score
1
July 18, 2009 8:50:41 PM

While we're on this topic, I've been wondering for a while why CD/DVD/etc. drives don't have multiple reading lenses. That would be the only way to make the read speeds faster since you can't get past 52x reading for CDs.
Having 2 reading lenses 1 or 2 tracks apart would double the reading speed, considering that reading from disks is mostly sequential. Just combine the data from the two lenses in a RAM buffer. And space is definitely not in issue in 5'1/4 drive.
Score
1
July 18, 2009 11:34:09 PM

13 years ago, Intel never knew that their 2009 batch of cpus would
be on a 32nm wafer . No one would come to think that vpu's would become
powerful GP-GPU, who would of thought that this cards would make
folding@home callculations faster than CPU's?

What most of you don't seem to realize (and that's maybe why enterprises
won't go with SSDs within the next decade) is one simple thing. In case
of an electrical outage, this babies will fry and ALL the data on them
will be lost for ever (or am I wrong?). That's why enterprises will go
with the conventional spin drives (Data on them can be restored and it
doesn't take much to do it!).

zodiacfml is right, RHD sounds way better for some one having an
enterprise, even if it comes in 5 years from now, it will sell. Data
restore would be easier to achieve from a RHD than a SSD.

That's my two cents.
Score
-2
July 19, 2009 3:39:10 AM

Instead of dual actuators, why not read/write magnetic domains (bits) from each platter simultaneously. E.g., Platter1/Head1 reads 2 bits and P1/H2 reads 2 and P2/H1 reads2 and P2/H2 reads 2...we have a full byte in 1/4th the normal rotation (yes my example is simplistic, I know about RLL, PRML, etc.). Only problem is a controller that is flexible enough to handle a variable numbers of platters. The marketing will take care of itself because the more platters/bigger capacity will be faster. This might allow traditional magnetic hard drives to remain competative with the onslaught of SDDs. Throw in dual-actuators and we would have a beast to contend with... ARE YOU LISTENING SEAGATE?
Score
0
July 19, 2009 8:40:51 AM

I always wondered why instead of using multi platters to increase capacity why not use the extra platter as an internal raid0 setup? I guess the increase for data faults would increase exponentially. It would effectively increase the width of data by the number of platters.
Score
0
July 19, 2009 8:41:28 AM

firehead2k Kenwood actually made a drive with two read optics.
Score
0
July 19, 2009 9:51:01 AM

Blessedmanfirehead2k Kenwood actually made a drive with two read optics.

Actually, if you talk about Kenwood's TrueX 72X CD-ROM, it only had one optical head, but it's laser was split into seven beams, which were processed in parallel, providing a lot more data throughput.

As for the dual-arm HDD's, I believe the biggest issue is that of storing servo information which can be used for both heads. The huge track density means an extreme precision is needed.
Also, using several heads in parallel is not possible because different platter will always dilate/contract slightly differently and the bits on each platter will not be always aligned.
Perhaps if each platter could have it's own arm, and all arms share the same axis, that would make possible to read several platters at the same time.
Score
0
July 19, 2009 10:57:22 AM

nekatrevenSo we just have to ask and the answer will be revealed?!?"Why do hotdog buns come in packs of 8 when hotdogs come in packs of 12?"(but really, nice article)


I've wanted this answer for years...

ie49589Very interesting article and nice pics of both the drives and the blue-prints. Not like the usual crap from Jane McEntegart.


Yes Tuan does exceptional professional articles, but Jane's article are still prettry cool. Sometimes you want to have a little personality. =D
I prefer a mixture of both, and a article like this fits Tuans writing style perfectly.

brandonvithis seems rather stupid honestly to me by the hard drive companys from what i understand a normal 7200rpm harddrive does somewhere around 75mb/s while the SATA 3.0 goes MUCH faster even if it cost $100 instead of $50 for a 500GB harddrive there are a lot of people that will pay that for 2x the speed and there are a lot that do by making raids i am right now planing to make a 4 drive raid in my next computer i have no need of more then even 500GB of total Harddrive space yet i am going to end up with 2TB of room instead buying 2 drives that were 2x as fast in a raid would of been somthing i would do with out a problem at all.


Interesting read, but please, proper sentence structure atleast a little.

tuannguyenThanks for the feedback guys!More articles like these coming. Hope everyone enjoys the weekend./ Tuan


Sure, if you count working overtime in a 40 degree plus shop...
But yeah, more articles like the technology inquisitive ones lately would be awesome. We're always learning, so may as well learn something cool.

ceterasActually, if you talk about Kenwood's TrueX 72X CD-ROM, it only had one optical head, but it's laser was split into seven beams, which were processed in parallel, providing a lot more data throughput.As for the dual-arm HDD's, I believe the biggest issue is that of storing servo information which can be used for both heads. The huge track density means an extreme precision is needed. Also, using several heads in parallel is not possible because different platter will always dilate/contract slightly differently and the bits on each platter will not be always aligned.Perhaps if each platter could have it's own arm, and all arms share the same axis, that would make possible to read several platters at the same time.


=D
As much as I do prefer single platter HDDs, a 4+ platter drive each running independantly in a raid setup (possibly something other than raid 0 for some data protection). It could run at 5k rpm rather than 15k rpm with a comparably throughput, potentially evening out for the increased failure rate that is destined by having more moving parts.


I'm told that. for atleast storage purposes, convential HDDs still have many years to come and are fart from reaching their physical limits, they may be a technology we see, assuming there is enough of a market for them to go for it.
Score
0
July 19, 2009 12:33:05 PM

Yeah, its RAID, thats what dual-actuators is. RAIIDDDDDDDD
Score
-1
July 19, 2009 4:30:37 PM

The Kenwood CD drives with multiple beam splitters were a good idea, but in reality just junk. My friend and I each got one and they failed quickly. They were both replace as part of the class action lawsuit that followed. Both of our replacement drives failed within a year. Sometimes riding the technology wave will bite you. Now I tend to let others prove the new tech ahead of me before I spend my hard earned dollars.
Score
0
July 19, 2009 4:40:53 PM

ceterasAlso, using several heads in parallel is not possible because different platter will always dilate/contract slightly differently and the bits on each platter will not be always aligned.

That explanation is irrelevant, since the data from each head would be buffered and merged independently for reads and split and synced independently for writes. Only the complexity of the controller would change... current hardware (actuators, heads and platters) would be compatible, as is.
Score
0
July 19, 2009 9:50:18 PM

I always wondered why no one made a hd such as this. It would be cheaper if one set of heads handled the inner tracks and the other the outer set, less logic needed in the controller. It would be similar to short stroking a current hard drive. It could even have a fall-back mode if one headset failed it could then use one set for the entire drive. They of course would have to be separate arms and not linked like Seagate's initial drive.

Score
0
July 20, 2009 2:45:38 AM

Just to be clear, I was not at all advocating using more than 1 head per platter side. I'm not against it either, since Tom's showed the advantages of short stroking big drives in an article a while back. I was amused that this article mentioned Seagate having used 2 heads per arm early on, which I somehow wasn't aware of (odd, since I own a very old copy of the 'Hard Drive Bible' from CSC that I thought I had thoroughly perused).
Score
0
July 20, 2009 5:36:27 AM

I remember years ago when dual-head was announced. Good grief, I'm really getting old :( .

One of the cited application is for a more secure access. Basically, one actuator has both read/write while the other is read-only. When a command is from a trusted user, the command is routed to the read/write actuator; otherwise it is routed to the read-only actuator.

I was wondering what happened to this tech/approach. Resetting timer.
Score
0
July 20, 2009 11:43:54 AM

Quote:
Seagate also mentioned to us that decades ago when harddrives were too heavy to even lift, many drives actually had sweep-heads--literally. These heads had camel-hair brushes on them that would sweep the platters clean as they turn. Such drives were utilized in large-scale mainframe systems.

That was random... and yesterday I ate toast for breakfast.
Score
0
July 20, 2009 1:31:19 PM

{While talking about the drive, one commenter proposed a question: why don't hard drive manufacturers do something like add a second set of read/write heads to increase performance rather than turning up the spindle rate to dangerously high levels?

Good question! We went directly to Seagate for the answers.}

Very nice. Our words can reach Seagate!

Now, somebody needs to tell seagate, that we are buying RAID setups.

We want dual head (and triple head). We want more platters (the simplest internal raid possible). we want short stroke a 15000 rpm 5.25" disk, because it would be faster that 3.25", meanwhile we will minimize head movement, but it does nothing for rotational latency. the only way to reduce rotational latency is multiple heads...

and Why the arm should be tangent to disk? a radial arm with many heads would move less?

extra heads would be able to read the disk if one fails.

We don't care about weigth, because 20% is ligter than 2 disks in RAID.
Score
0
July 20, 2009 1:33:12 PM

firehead2kWhile we're on this topic, I've been wondering for a while why CD/DVD/etc. drives don't have multiple reading lenses. That would be the only way to make the read speeds faster since you can't get past 52x reading for CDs.Having 2 reading lenses 1 or 2 tracks apart would double the reading speed, considering that reading from disks is mostly sequential. Just combine the data from the two lenses in a RAM buffer. And space is definitely not in issue in 5'1/4 drive.


I think that faster rotational speed would damage the disks, but if so, then is possible to rotate also the lens in opossing direction, increasing relative speed.
Score
0
July 20, 2009 2:34:17 PM

tuannguyenBecause Seagate says it's 15,000 RPM?http://www.seagate.com/docs/pdf/ma [...] _15k_7.pdfAnd the ".7" in 15K.7 doesn't represent "15,700" RPM./ Tuan



It actually represents the 7th generation of cheetah drives, which is why the last cheetah was called 15K.6, 15K.5, 15K.4 and so forth.
Score
0
July 20, 2009 4:19:30 PM

By having multiple platters, aren't there multiple actuators as well?

They made it vertical instead of horizontal. elegant.

With densities as they are, I can see temperature fluctuation being a real problem.

RAID is another means of adding actuators, as a previous poster stated. Very astute.

My question is not with hard drives (wehere contact between the disk and head exists). My questions is why not use a set of mirrors for optical media. This concept was first executed by the Soviet Airforce in the MiG 21 durring the late 70's with a RADAR dish that deflected the beam producing surface, rather than having a rotating traditional dish. It gave a better view of the area with a faster sweep because the dish basically had only 2 positions. I could see an adaptation for the optical drive that could increase throughput from a few megabytes per second to several hundred megabytes per second.
Score
0
Anonymous
July 20, 2009 5:17:14 PM

Cost factors aside, Why not take SSD technology and combine it with todays spindle technology to increase over all system performance for functions like Paging. Setup interupts in the bios to acknoledge disks with high speed ssd Boot and Paging device functions. Have the OS boot from the first half of the SSD, use the second have for virtual memory paging and store high volume data on the spindle portion of the disk. Since SSD's produce very little heat, the drives operating tempurature wouldn't need to be great consideration for overal drive life.

The only draw back would be that the Paging portion of the SSD may burn out faster than the rest of the disk, but most virtual memory page files rarely exceed 8GB. (in my humble experience) So even a 32GB Paging apparatus could provide 4 times the normal wear and tear of a drive being soley used for a complete data storage device.

Well, theres my 2 bits.
Score
2
July 20, 2009 5:38:05 PM

...
Score
-2
!