Microsoft Repositions 7TB 'Project Silica' Glass Media as a Cloud Storage Solution

News
By Mark Tyson
published

7 TB glass storage media is now faster to write and "could last 10,000 years."

Project Silica
(Image credit: Microsoft Research)

Microsoft has provided an update on its Project Silica storage research. According to a recent blog post and video, the software and devices giant is now positioning Project Silica as a sustainable storage solution to address the cloud storage server market. Using Project Silica technology, it is possible to store approximately 1.75 million songs or around 3,500 movies on a palm-sized slice of glass. Moreover, Microsoft asserts that the 7 TB storage per glass sheet maintains data integrity for 10,000 years.

Several refinements have been made to Project Silica since our last report in June 2022. At that time, the media seemed identical in dimensions, but the writing process has since been refined for speed and durability. Also, we note the focus seems to have changed from music archiving purposes to cloud server data storage.

In the new video, Ant Rowstron, a deputy lab director at Microsoft Research Cambridge, outlines the great appeal of Project Silica in addressing the ever-increasing amount of data being generated by humans and in cloud storage.

Sketching out the limitations of magnetic storage, a traditional cloud server data medium, Rowstron highlights the need to refresh the data every five years when using HDDs and up to ten years “if you are brave” with tape storage. Refreshing media can be arduous, wastes energy, and introduces the potential for data corruption.

In contrast, data stored in glass, with its natural resistance to water, electromagnetic pulses, extreme temperatures, and surface scratching, can be stable for thousands of years.

The video also takes us through the four steps to silicon storage heaven. There are four ‘labs’ involved in Project Silica data storage and retrieval, as follows:

  1. Write Lab: short laser pulses record data inside the glass as voxels (3D pixels)
  2. Read Lab: a computer controlled microscope reads data
  3. Decode Lab: here the data read is decoded back into standard computer readable formats using Azure AI
  4. Library lab: when data is requested for recall, a robot in the library goes to fetch the appropriate piece, and inserts it into a reader.

(Image credit: Microsoft Research)

It is emphasized that the library is immutable and passive. All the complexity and power in the system is in the robot. We wondered what would happen if a robot fell or dropped a slice of glass, but Microsoft’s blog doesn’t discuss this possibility.

Image 1 of 3
Project Silica
(Image credit: Microsoft Research )

Last but not least, it is interesting to note Microsoft’s change of direction and some quite incredible claims for Project Silica. The firm admits it still isn’t ready for commercial use, though. It envisions glass storage as “a mainstay in Azure data centers across the globe” that needs another three of four developmental stages to secure its place and live up to its durable, sustainable, and cost-effective potential.

Mark Tyson
Mark Tyson
Freelance News Writer

Mark Tyson is a Freelance News Writer at Tom's Hardware US. He enjoys covering the full breadth of PC tech; from business and semiconductor design to products approaching the edge of reason.

27 Comments Comment from the forums
  • DougMcC
    Drop a piece of glass? Why would you design a system where that was physcially possible? Relatively trivial to design a pull-only system where the arm pulls the glass from a storage tray, and gravity makes it physically impossible for the glass to go anywhere but the pull arm. Then the pull arm delivers the glass to the reader which again pulls the glass onto a read device. There's never a need for it to be possible for the glass to fall.

    However, breakage is not impossible (earthquake, asteroid hit) which is why all such systems are also designed with redundancies of various strategies, including off-site backups for that asteroid hit.
    Reply
  • Kamen Rider Blade
    But when will this type of technology become consumer ready?

    I can see them using DLP like Mirrors to read/write across a volume of Glass.

    Similar to the Optolythic Data Rod or Isolinear Rods that was used in Star Trek by the Cardassians for WORM (Write-Once Read Many) type storage.
    Reply
  • usertests
    Kamen Rider Blade said:
    But when will this type of technology become consumer ready?

    I can see them using DLP like Mirrors to read/write across a volume of Glass.

    Similar to the Optolythic Data Rod or Isolinear Rods that was used in Star Trek by the Cardassians for WORM (Write-Once Read Many) type storage.
    As described with the stationary slab and complicated write/read equipment, consumers are not getting it. Project Silica seems to be similar to other technologies like the "5D optical / Superman memory crystal".

    The industry is reluctant to make a new type of spinning 12 cm optical disc or "hard drive" for consumers. Write-once is certainly less convenient, but ordinary people would definitely do it if they could archive terabytes or hundreds of terabytes with indefinite longevity. It would be nice to store photos, videos, whatever without the relatively high risk of data loss associated with HDDs, SSDs, and CD/DVD/Blu-ray (20-50 years). M-DISC might be the only acceptable option right now, if it works as advertised, but its $/TB is awful.
    Reply
  • rluker5
    usertests said:
    As described with the stationary slab and complicated write/read equipment, consumers are not getting it. Project Silica seems to be similar to other technologies like the "5D optical / Superman memory crystal".

    The industry is reluctant to make a new type of spinning 12 cm optical disc or "hard drive" for consumers. Write-once is certainly less convenient, but ordinary people would definitely do it if they could archive terabytes or hundreds of terabytes with indefinite longevity. It would be nice to store photos, videos, whatever without the relatively high risk of data loss associated with HDDs, SSDs, and CD/DVD/Blu-ray (20-50 years). M-DISC might be the only acceptable option right now, if it works as advertised, but its $/TB is awful.
    I archive on m-disc, but I doubt it will last long.
    Blu-ray players are practically obsolete.

    In 5 years they will have gone the way of the cassette tape(which I also used to use for data), the 8-trac and the wire recorder whatever they were called.

    I wish they would stay in use, but I doubt they will.
    For example I bought my m-disc burner in 2013.
    It does everything but litescribe, even copies uhd since it was made pre drm. No innovation in 10 years tells me the tech is dead. The only Blu-rays I use anymore are the occasional 3d one for entertainment.

    My m-discs will only be readable as long as I can get a working player. It isn't a 1000 year storage option and neither is rare pieces of glass.
    Reply
  • Kamen Rider Blade
    rluker5 said:
    I archive on m-disc, but I doubt it will last long.
    Blu-ray players are practically obsolete.

    In 5 years they will have gone the way of the cassette tape(which I also used to use for data), the 8-trac and the wire recorder whatever they were called.

    I wish they would stay in use, but I doubt they will.
    For example I bought my m-disc burner in 2013.
    It does everything but litescribe, even copies uhd since it was made pre drm. No innovation in 10 years tells me the tech is dead. The only Blu-rays I use anymore are the occasional 3d one for entertainment.

    My m-discs will only be readable as long as I can get a working player. It isn't a 1000 year storage option and neither is rare pieces of glass.
    That's why I'm pissed at Sony / Panasonic for not focusing | Educating | Building on consumer archival technology.

    They have Archival Disc sitting in their Enterprise Line-Up.

    Zero Intent to bring it to consumers for us to benefit.
    Reply
  • usertests
    rluker5 said:
    My m-discs will only be readable as long as I can get a working player. It isn't a 1000 year storage option and neither is rare pieces of glass.
    The people who actually care about 1000 or billion year longevity are probably adding pictograph instructions for post-apocalypse humans or aliens to look at, as long as they can bring their own microscope. It might be easier to reverse engineer a reader that can produce the correct sequence of binary bits, than it would be to correctly interpret the data formats used, especially if there is a species barrier.

    We can still use this kind of longevity in the consumer market, if it effectively removes degradation of the storage medium as a concern. You would prefer if your storage didn't fail in under 50 years. Yet there are plenty of HDD, SSDs, and optical discs that have failed in under 5 years from mishandling or bad luck.

    CDs have been around for 41 years, and there are still relatively cheap drives being mass produced that can read CDs. If we had another 12 cm consumer disc to succeed Blu-ray, the demise of CD/DVD/Blu-ray readability could be postponed since most drives maintain backwards compatibility (notably, the PlayStation 5 doesn't support CDs while the Xbox Series X does). But that ain't happening and the disc drives continue to be dropped from consumer electronics. We're left with other options with their own problems, like big HDDs that can fail suddenly and catastrophically, or SSDs using ever-worse NAND in pursuit of density (PLC will be on the menu soon).
    Reply
  • rluker5
    Kamen Rider Blade said:
    That's why I'm pissed at Sony / Panasonic for not focusing | Educating | Building on consumer archival technology.

    They have Archival Disc sitting in their Enterprise Line-Up.

    Zero Intent to bring it to consumers for us to benefit.
    Archival Disc would have been an advancement.
    But the time it takes to read and write on optical discs is pretty inconvenient. It is hard to get past that.
    Reply
  • lmcnabney
    I highly suspect the exaggerated lifespan is BS. Glass isn't really a solid - it flows exceptionally slowly. Yeah, it take a long time. After a hundred years the top of a pane of glass is visibly thinner than the bottom. This very slow flow is inconsequential for almost any usage, but when storing very tiny bits ANY movement is bad and as the decades pass there will be movement.
    Reply
  • Kamen Rider Blade
    usertests said:
    CDs have been around for 41 years, and there are still relatively cheap drives being mass produced that can read CDs. If we had another 12 cm consumer disc to succeed Blu-ray, the demise of CD/DVD/Blu-ray readability could be postponed since most drives maintain backwards compatibility (notably, the PlayStation 5 doesn't support CDs while the Xbox Series X does). But that ain't happening and the disc drives continue to be dropped from consumer electronics. We're left with other options with their own problems, like big HDDs that can fail suddenly and catastrophically, or SSDs using ever-worse NAND in pursuit of density (PLC will be on the menu soon).
    "Archival Disc" was supposed to be the successor to "Blu-Ray" as a WORM storage medium.

    If you take the current "Archival Disc" density of 83.3… GB/layer
    A single "Quad Layer" disc could net you 333 GB using the existing Blu-Ray Quad-Layer stack near the bottom.

    If you tried to go for the hybrid 3-layer Stack in the middle and 4-layer Stack on the outside that the HD-DVD consortium proposed to hybridize HD-DVD & Blu-Ray, that would allow for a cheap 7-Layer per side Optical Disc.
    If you used a Mini-Disc like Cartridge system, I came up with a specification for a 130mm Optical Disc that would net you 40% more Optical Data Area over the 120 mm standard Optical Disc by Min/Maxing the inner & outter areas and using Sonys existing smaller Metal hub.

    40% more Data area for 10mm more Optical Disc Diameter is a "BIG DEAL" if you're willing to break backwards compatibility and create a dedicated "Double-Sided" Archival Disc using a Mini-Disc style 4mm thick Protective Cartridge system to protect the "Double-Sided" 7-Layer Optical Disc.

    At 7-Layers I estimated a data capacity of 1520.506 GiB on one side spread across 7-Layers.

    3041.012 GiB across 2-sides.

    But with the properly designed Optical Disc Drive, you can Read/Write to both sides simultaneously.

    Especially if you're using Multi-X like technology.

    rluker5 said:
    Archival Disc would have been an advancement.
    But the time it takes to read and write on optical discs is pretty inconvenient. It is hard to get past that.
    The old Kenwood 72x CD-ROM technology where they split 1x beam into 7x beams and read back the data was legit. I owned it, it was great.
    If Kenwood kept updating it with the times as Optical Disc Mediums evolved, it would've gotten MUCH faster and made reading a entire disc relatively painless.

    They already had 7x Beams working in the commercial 72x variant.
    They were doing R&D on splitting into 11 beams.
    One day, I was hoping they would hit 15x different beams.

    I've estimated how fast a modern 7x Beam Multi-X type drive would perform on Blu-Ray:
    For Marketing / TradeMark purposes, I called it a HeptaDrive, but it's still the Multi-X drive technology that Kenwood pioneered all those decades ago.

    I just wish Sony would take the ball and update it.
    Reply
  • usertests
    Kamen Rider Blade said:
    "Archival Disc" was supposed to be the successor to "Blu-Ray" as a WORM storage medium.

    If you take the current "Archival Disc" density of 83.3… GB/layer
    A single "Quad Layer" disc could net you 333 GB using the existing Blu-Ray Quad-Layer stack near the bottom.
    I'm aware of Archival Disc. I don't think they're priced for consumers, and I'm not sure if the 1 TB version has appeared in any form. Despite 1 TB optical discs being demoed in the lab over a decade ago.

    The last chance for consumers would be a new capacity/drive push for 8K content, which doesn't seem like it will happen for cost reasons, and because streaming services can use DRM more effectively.
    Reply