Sign in with
Sign up | Sign in

Analysts Predict Acceleration in HDD Storage Density Growth

By - Source: IHS | B 29 comments

Market research firm IHS predicts a fast increase in HDD storage densities and, as a result, a similarly strong increase in storage capacities.

The faster pace will be driven by a need to record massive amounts of audio and video data and require a new technology that can break through the density barrier of the current perpendicular magnetic recording (PMR) technology.

IHS believes that the average density of 744 Gb per square inch in 2011 will grow at an average of about 19 percent over the next few years and reach 1.8 Tb by 2016. 2012 will see a density of 780 Gb per square inch per platter, and then rise to 900 Gb per square inch next year, the market research firm forecasts.

The trend calls for a technology that will allow an increase of the number of bits that can be safely stored on a platter without interfering with each other. The density improvements of storage densities have been nothing short of amazing over the past 15 years. In 1997, Seagate acquired Quinta, a developer of heat-assisted magnetic recording technology that was believed to drive the then believed physical limited of about 100 Gb per square inch to about 250 Gb per square inch using its OAW (Optically Assisted Winchester) technology.

Manufacturing improvements and especially the transition from longitudinal recording to PMR delayed the introduction of the rather expensive production of heat-assisted magnetic recording drives and gave the industry an opportunity to approach 1 Tb per square inch. OAW, however, has been in development as HAMR (heat-assisted magnetic recording) technology at Seagate and has already reached 1 Tb per square inch in a lab environment. If IHS is correct, we can expect the arrival of this technology in commercial drives by 2014.

Display 29 Comments.
This thread is closed for comments
Top Comments
  • 23 Hide
    mman74 , May 22, 2012 1:48 AM
    ... and increasing transfer speeds?
  • 21 Hide
    mman74 , May 22, 2012 1:47 AM
    What about reducing the cost per TB?
  • 18 Hide
    igot1forya , May 22, 2012 2:03 AM
    I still think they need to improve the mechanical actuator technology - virtually no innovation has advanced this field (at least that I have heard). Start by making the actuator fixed and instead of a small read/write sensor on a small head, replace it with a solid read/write strip along the diagonal surface of the drive, then increase the number of strips to boost IO performance even further. But alas, I'm just a dreamer and not a storage engineer.
Other Comments
  • 12 Hide
    drwho1 , May 22, 2012 1:42 AM
    Where are those 60TB hard drives?
  • 21 Hide
    mman74 , May 22, 2012 1:47 AM
    What about reducing the cost per TB?
  • 23 Hide
    mman74 , May 22, 2012 1:48 AM
    ... and increasing transfer speeds?
  • 14 Hide
    sabot00 , May 22, 2012 1:51 AM
    Well, increased density is basically linearly increased speed.
  • 18 Hide
    igot1forya , May 22, 2012 2:03 AM
    I still think they need to improve the mechanical actuator technology - virtually no innovation has advanced this field (at least that I have heard). Start by making the actuator fixed and instead of a small read/write sensor on a small head, replace it with a solid read/write strip along the diagonal surface of the drive, then increase the number of strips to boost IO performance even further. But alas, I'm just a dreamer and not a storage engineer.
  • 2 Hide
    Plasmid , May 22, 2012 2:14 AM
    It all about ssds at this age!!!
  • 9 Hide
    TheBigTroll , May 22, 2012 2:29 AM
    seagate should make a 10k rpm consumer drive like WD but this time, throw in a 16gb SSD cache. this would be killer!!
  • 8 Hide
    icrf , May 22, 2012 2:32 AM
    igot1foryaI still think they need to improve the mechanical actuator technology - virtually no innovation has advanced this field (at least that I have heard). Start by making the actuator fixed and instead of a small read/write sensor on a small head, replace it with a solid read/write strip along the diagonal surface of the drive, then increase the number of strips to boost IO performance even further. But alas, I'm just a dreamer and not a storage engineer.

    The problem with a static strip of read/write heads is that the head is much much larger than the track it is reading. There's no way to make them dense enough to cover the platter, so they have to move. There may be benefit had in placing more than one head on a single actuator on the same side of one platter.
  • 5 Hide
    rohitbaran , May 22, 2012 2:42 AM
    A few days back, some 'researcher' talks about the SSD's bleak future and now this. Do I smell a proxy campaign?
  • 6 Hide
    TheBigTroll , May 22, 2012 2:48 AM
    rohitbaran: that was like a month ago
  • 1 Hide
    blazorthon , May 22, 2012 3:01 AM
    sabot00Well, increased density is basically linearly increased speed.


    Not really. Data density (there's probably a more proper term for that) is increased in hard drives by increasing the amount of tracks and the storage per track. Only increasing the storage per track improves sequential performance, so as density increases exponentially, performance then increases more linearly (assuming that track count and data per track increase at similar rates).

    That would be why lower capacity drives, such as a 500GB drive, are not several times slower than a much higher capacity drive with the same spindle speed.
  • 3 Hide
    CaedenV , May 22, 2012 4:20 AM
    blazorthonThat would be why lower capacity drives, such as a 500GB drive, are not several times slower than a much higher capacity drive with the same spindle speed.

    Almost; The reason why a 500GB drive is of a similar speed of much larger drives is due to the number of platters. A 500GB and a 4TB drive may be identicle with the excetion of how many platters are used (though higher end drives use more platters of lower density for more edge space). Because consumer drives only read a single head/platter at a time, the speed remains the same (or at least very similar) no matter the size of the drive, but the density of the platters used. Other than that you are spot on.
    igot1foryaI still think they need to improve the mechanical actuator technology - virtually no innovation has advanced this field (at least that I have heard). Start by making the actuator fixed and instead of a small read/write sensor on a small head, replace it with a solid read/write strip along the diagonal surface of the drive, then increase the number of strips to boost IO performance even further. But alas, I'm just a dreamer and not a storage engineer.

    I completely agree! While what you speak of is science fiction, you are on to something; we need more read heads active at the same time! Borrow tech from the SCSI world, where drives can read and write to multiple platters at the same time, or have multiple actuators that work in conjunction with eachother (like having a dedicated actuator per platter/side, or dedicated inner/outer actuators). HDDs have plenty of sequential performance, but they simply need something more than a single arm or head being used in order to get that seek time down, and random access time up!

    @article
    This is not news! This is a speculation company making speculations and has 0 founding on actual science, technology, or product announcements. It is all well and good for them to think there will be an acceleration in data storage for HDDs, but really it is a moot point. 4TB drives are out now, and have more than enough space for most users. 6TB drives will be released in a year or so, and a pair of those should make even the most insane home users very happy. Meanwhile in server world they are finally making it to 2TB drives, and most are still installing drives in the 300-750GB range. Then we have products like the R4 16TB SSD that takes roughly the space of 2 HDDs in the form of a card. Much less heat, much less power, much less noise, much less space, insane amounts of performance, and (after figuring in the costs of maintenance and use) it is roughly the same price over its expected life as a mess of SAS HDDs.
    Quite frankly; HDDs are dead for system drives, and they are dead for server applications. It is just a matter of time before it is a dead tech for home storage as well. The whole tech is going to burn very quickly, with SSDs more capable of higher densities due to stack-ability (seriously, platters are fat! not to mention all the support hardware of the acoustic housing, and actuator) and heat dissipation which are 2 things that HDDs can never combat effectively.
  • 1 Hide
    blazorthon , May 22, 2012 4:34 AM
    CaedenVAlmost; The reason why a 500GB drive is of a similar speed of much larger drives is due to the number of platters. A 500GB and a 4TB drive may be identicle with the excetion of how many platters are used (though higher end drives use more platters of lower density for more edge space). Because consumer drives only read a single head/platter at a time, the speed remains the same (or at least very similar) no matter the size of the drive, but the density of the platters used. Other than that you are spot on.I completely agree! While what you speak of is science fiction, you are on to something; we need more read heads active at the same time! Borrow tech from the SCSI world, where drives can read and write to multiple platters at the same time, or have multiple actuators that work in conjunction with eachother (like having a dedicated actuator per platter/side, or dedicated inner/outer actuators). HDDs have plenty of sequential performance, but they simply need something more than a single arm or head being used in order to get that seek time down, and random access time up!@articleThis is not news! This is a speculation company making speculations and has 0 founding on actual science, technology, or product announcements. It is all well and good for them to think there will be an acceleration in data storage for HDDs, but really it is a moot point. 4TB drives are out now, and have more than enough space for most users. 6TB drives will be released in a year or so, and a pair of those should make even the most insane home users very happy. Meanwhile in server world they are finally making it to 2TB drives, and most are still installing drives in the 300-750GB range. Then we have products like the R4 16TB SSD that takes roughly the space of 2 HDDs in the form of a card. Much less heat, much less power, much less noise, much less space, insane amounts of performance, and (after figuring in the costs of maintenance and use) it is roughly the same price over its expected life as a mess of SAS HDDs.Quite frankly; HDDs are dead for system drives, and they are dead for server applications. It is just a matter of time before it is a dead tech for home storage as well. The whole tech is going to burn very quickly, with SSDs more capable of higher densities due to stack-ability (seriously, platters are fat! not to mention all the support hardware of the acoustic housing, and actuator) and heat dissipation which are 2 things that HDDs can never combat effectively.


    I meant to include platter count in that statement too. Both are factors as to why lower capcity drives are almost as fast as higher capacity drives (or just as fast).
  • 0 Hide
    alidan , May 22, 2012 4:43 AM
    icrfThe problem with a static strip of read/write heads is that the head is much much larger than the track it is reading. There's no way to make them dense enough to cover the platter, so they have to move. There may be benefit had in placing more than one head on a single actuator on the same side of one platter.

    IDDDDDEEEEEEEEEAAAAAAA!!!!!!!!

    how about instead of a static strip, or just one on the head, there could be 5 or 6 up and down on a movable body.

    when the harddrive is in well working order, it would read at least 5 times faster than current, and because there are multipul heads, would also probably reduce the seak time.

    the only problem would come in the form of a bad sector and data needing to be moved, but even than, it would just fall back to current speeds.

    best case (first year or so) performance is great, after a while, performance will degrade without maintenance and ageing hardware.
  • 0 Hide
    blazorthon , May 22, 2012 4:56 AM
    alidanIDDDDDEEEEEEEEEAAAAAAA!!!!!!!!how about instead of a static strip, or just one on the head, there could be 5 or 6 up and down on a movable body.when the harddrive is in well working order, it would read at least 5 times faster than current, and because there are multipul heads, would also probably reduce the seak time. the only problem would come in the form of a bad sector and data needing to be moved, but even than, it would just fall back to current speeds.best case (first year or so) performance is great, after a while, performance will degrade without maintenance and ageing hardware.


    Maybe you could just squeeze four heads per surface (four actuators, one on each of four sides of the platters). That could increase both sequential and random access time greatly. Even if no one task used all of the performance at once, it could do up to four read/write operations at once, so despite it's still lower than SSD random access performance, it could be able to do multiple sequential jobs at once and still have great performance on all of them.

    Being able to do things such as running an AV scan, a defrag, decompressing a large archive, and playing a game all at once would be quite the job for a multi-tasking system without an SSD. Something like this could be priced between SSDs and HDDs while delivering performance that for most people might as well be from an SSD because it does the job more than well enough already. It wouldn't be able to touch SSDs for huge random read/write performance, but it would do the trick for multi-tasking and high enough sequential performance while having pretty good random access performance.

    Buy it for up to 25 cents or so per GB and it'd be a great purchase.
  • 0 Hide
    danwat1234 , May 22, 2012 5:27 AM
    19% density increase per year? How is that fast or faster than what we have seen previously? Once 2TB hard drives came out, it slowed down. The transition from 500GB being the largest drive to being able to buy to a 2TB 7200RPM drive was only like 4 years. 4 years for a 4x density increase. We won't be seeing this type of growth speed again or at least not in the near future..
  • 2 Hide
    danwat1234 , May 22, 2012 5:29 AM
    sabot00Well, increased density is basically linearly increased speed.

    It's actually exponential because the speed builds on itself because sequential read/write speed is directly related to track density and density increases exponentially (although recently it has slowed down so much that I can't call it exponential anymore). But as Blazorthon said, the curve for increasing speed is slower than how fast density increases, because density increases have to do both with track density and how many tracks per inch. I'd guesstimate about every 5-6x of density is double the sequential read/write speed.
  • 1 Hide
    ojas , May 22, 2012 5:57 AM
    Meh...where's my MRAM SSD :p 
  • 1 Hide
    billyboy999 , May 22, 2012 6:57 AM
    alidanIDDDDDEEEEEEEEEAAAAAAA!!!!!!!!how about instead of a static strip, or just one on the head, there could be 5 or 6 up and down on a movable body.when the harddrive is in well working order, it would read at least 5 times faster than current, and because there are multipul heads, would also probably reduce the seak time. the only problem would come in the form of a bad sector and data needing to be moved, but even than, it would just fall back to current speeds.best case (first year or so) performance is great, after a while, performance will degrade without maintenance and ageing hardware.


    Multiple independent read-write heads has generally been written off as impractical in terms of cost and complexity. You can do similarly well by writing your software better, installing more RAM, using an SSD, or using multiple drives (RAID).

    In a hard drive, there are 2 heads per platter - one on the top, one on the bottom. So in a, say, 3 platter hard drive, there are actually 6 R/W heads. Now, we know there's only one actuator for all these arms the heads are on, so they all have to move at the same time. You might think that during R/W operations, all heads are active, writing or reading in parallel from each of their platters. Fun fact: At any time only one of the heads are active. It turns out it's way too hard to get all of the heads to read/write at the same time (which would lead to incredible linear R/W speeds on multi-platter drives) due to the fact that the tiniest shift would result in a head not being aligned enough to read its data.

    So think about it - if hard drive companies have decided they can't even achieve this kind of parallelism, going for more than one set of actuators/arms would asking for the impossible.
  • 0 Hide
    blazorthon , May 22, 2012 7:02 AM
    billyboy999Multiple independent read-write heads has generally been written off as impractical in terms of cost and complexity. You can do similarly well by writing your software better, installing more RAM, using an SSD, or using multiple drives (RAID). In a hard drive, there are 2 heads per platter - one on the top, one on the bottom. So in a, say, 3 platter hard drive, there are actually 6 R/W heads. Now, we know there's only one actuator for all these arms the heads are on, so they all have to move at the same time. You might think that during R/W operations, all heads are active, writing or reading in parallel from each of their platters. Fun fact: At any time only one of the heads are active. It turns out it's way too hard to get all of the heads to read/write at the same time (which would lead to incredible linear R/W speeds on multi-platter drives) due to the fact that the tiniest shift would result in a head not being aligned enough to read its data. So think about it - if hard drive companies have decided they can't even achieve this kind of parallelism, going for more than one set of actuators/arms would asking for the impossible.


    I have to disagree. It would be harder to have a single actuator read/writing to multiple spots in a set of platters unless either it was read/writing to tracks on different platters all within the same cylinder or if each head on the actuator was able to move independently of the other heads on the same actuator without any one head shifting the others when any one head moves. Having multiple actuators should improve performance where multiple heads per actuator do not help much at all because multiple actuators would not have the aforementioned problem because they would be independent and spaced farther apart so as to not interfere with each other.
Display more comments