Google glasses are out in developers hands to fuel the creation of applications and the first information are starting to leak out.

A wink is all it takes

One of this application, called “Winky” for the time being, wants to enable the life logging. Life logging is a way to record our life, every moment, in a digital form. Storage is no longer an issue since we can have TB of space for less than 100€, and getting cheaper as we speak. In the coming years storage may even become irrelevant as the network provides ubiquitous and unlimited capacity thus de-localizing the storage from the point of use of data.

It remains the problem of capturing the “snapshot” of our life that we want to record in a seamless way. There have been a number of proposals in this field, like the Memoto automatic lifelogging camera on sale for 279$ including any required Cloud Storage, but the problem is that these systems capture “too much”.

Of course you can lifelog by using your cell phone, and matter of fact many youngsters are doing just that by clicking and clicking and posting on twitter, Facebook and more. But it is not always convenient to pick up your phone, frame, click and post.

Here it is where Winky comes handy: provide a seamless way to capture moments of your life by … winking. It exploits the Google glass camera and its capability to recognise the eyelids movement. You wink, a photo is taken on what you are looking at.

As it gets easier to capture the world around us it gets weirder to live in such a world. We already have hundreds of security cameras taking note of what we do but at least these cameras are supposed to be regulated (not 100% true, since I doubt very much that private security cameras collected data comply to any regulation…).

In the near future we might have (we will have) thousands of cameras recording photos in which we will be part, although in most cases we will be completely unaware. Are these photos going to hunt us in the future by bringing us back to an awkward past?

Current storage technologies are based on materials that can change its characteristics (like magnetisation) from one state to another. By attributing a value (0 or 1) to a specific state we can use it to “store” that bit. Changing the state means changing the value of the retained value. And, of course, by reading the state we “read” the value.

There are many technologies that allows us to do just that and the writing and reading occurs via the displacement of electrons. To have a reliable writing and reading, intuitively, you need to have many electrons moving around since they are fleeting and you cannot trust just one of them, nor hundreds…. It follows that you need to have many atoms in the substrate for storing just one bit (many today means in the order of 2 thousands atoms).

Configuration of a resistive storage cell (ReRAM): An electric voltage is built up between the two electrodes so that the storage cells can be regarded as tiny batteries. Filaments formed by deposits during operation may modify the battery’s properties. (Credit: Jülich Aachen Research Alliance (JARA))

Now scientists are learning to build up materials from the bottom up and in this way they can create substances having some desired properties. By applying nano tech a team of scientists of  Jülich Aachen Research Alliance (JARA) have been able to create a storage cell working on ions, rather than on electrons (by the way, living things also use ions, not electrons for their electrical communications). Ions are bigger than electrons, thousands time bigger (a single proton mass is almost 2000 times “bigger” than an electron – if you are picky the real ratio is 1836.152 672 45), and therefore can be controlled much better.

Don’t be misled: using ions doesn’t mean that you have to have bigger cells. When using electrons, implicitly you have to use also the atoms that have those electrons, so you are using electrons and ions…. If you use ions for storing and retrieving information you can have a reliable system with just a few of them. Hence the potential squeeze of dimension and the increase in storage density. In the figure you can se the schematics of the storage cell developed by the team. The size of the cell is 10nm so you can fit 10 billion of them in a single 1 square mm. A cubic mm of these cells would be able to store 1 EB of data (a million TB)!

Of course you cannot pack these cells side by side, and you need connectors so that eventually one could imagine having ONLY one PB (a thousands of TB) in a cubic millimetre. Not bad, though!

Now, we are nowhere near an industrial manufacturing of this kind of storage and what they have is just a paper on the Journal Nature Communications reporting the result of a prototype they manage to build of a single cell.

Even though we are far away from a commercial application, remember that 70 years ago the first transistor was a bulky piece of germanium …

I know, as you, the three words but seeing them together puzzled me. Quantum properties connected to magnetisms via Femto (10 to the minus 15!) looks intriguing.

A laser pulse, so much shorter than a blink of an eye!

This triplet appears in a paper just published on Nature, where Researchers at the U.S. Department of Energy’s Ames Laboratory, Iowa State University, and the University of Crete in Greece report on a successful experiment to change the magnetisation of a special material, having what is know as Colossal MagnetoResistive characteristics (CMR), in an incredibly short time: 100 femtoseconds.

This is the time they apply a laser pulse, a time in which light travels for just 3 cm!

Today laser is being used to heat up the area where a bit is being stored to decrease the amount of magnetic field that is required to change the magnetisation of the material (HAMR – Heated Assisted Magnetic Recording).  In turn this makes it possible to have higher density storage.

This experiment aims at increasing the speed of recording, and indeed it can lead to a 1,000 fold increase in speed. That means to be able to write Tbps rather than Gbps as it is today.

The experiment shows that it can indeed be done, but scientists still need to understand “why”! The fact is that at this level of time window the effects that take place can no longer be explained in terms of classical physics. One has to turn to quantum mechanics and see what happens not to atoms but to electron’s spin. And this is how the quantum-femto-magnetism name comes up.

The concrete application of this phenomenon is still far down the lane, but clearly it shows that spintronics is now starting to move from being a theoretical possibility to being an issues that needs to be solved by engineers to make it industrially and economically viable.

A nano

When you play a guitar each of the six strings “stores” one note. Well, researchers at the Technical University of Munich are looking at using the vibration of a string to store information, more specifically they are trying to use a string made by a single  carbon nanotube to store “quantum” values from 0 to 1.

A carbon nanotube is held at the two extremities and an electromagnetic field makes it vibrate. The nanotube is so small that the vibrations are of the order of millions per second (the shorter the string the higher the frequency…). It can vibrate in many directions and these directions can be measured by a laser beam. It can actually be constrained to vibrate in two specific directions and hence one could associate the value of one to a direction and the value of zero to the other. Or one can relax the constraints and let it vibrates in the 3D space hence potentially assuming any value between 0 and 1.

This mechanism, therefore, can become a component of a quantum computer, able to store an information for as long as a second. It does not seem a very long span of time, but we are talking about quantum phenomena that happens at the nanosecond scale. So as a transient memory used to perform computation one second is long enough.

A microwave cavity resonator

Almost simultaneously, researchers from Yale have published a paper on Nature describing a different approach to store information in a quantum computer. They are proposing to use photons (that interact weakly and seldom with anything else) to store information in a quantum computer.

They have managed to change the state of a photon using a microwave cavity resonator as a medium. Since photons do not interact (basically) with anything else their status, and hence the information associated, can be preserved for a long time.

Another piece is the quantum computing puzzle that anyhow remains at the end of the rainbow. As we make one step forward the end of the rainbow with its pot of gold backtrack one step….

It is not possible yet to fix a date for the first usable running quantum computer.

The RAMAC 305 was so big and heavy that a forklift was needed to move it around (credit IBM).

No it is not a misprint. This is what IBM scientists said at the launch of their RAMAC 305 in September 1956. That first hard drive could store 5 MB and scientists said that the technology used would have allowed a hard disk with a bigger capacity but who would ever need more than that?

It didn’t come cheap either<. in today’s dollar it would cost 160,000$, that is 32,000$ per MB. Today the cost for 1MB of storage on a hard disk is about 0.0001$. And although you can no longer find a hard disk storing 5MB you can easily fit a 100$, 1TB hard disk in your pocket. You would need 200,000 of RAMC 305 to store 1TB …

It is amazing to think how much scientists have been able to squeeze the space needed to store bits. Lately they have invented a way to heat the magnetic substrate to be able to use a weaker magnetic force to store a bit so that one can use a smaller area without affecting nearby areas (HAMR, Heat Assisted Magnetic Recording).

However heating is expensive in terms of power consumption and can only go as far. Here comes a new approach to storage invented by some researchers at the Oregon State University.

They have found that it is possible to focus a tiny beam of ultrasound waves, hitting a microscopic area of the magnetic surface. This wave deforms the surface making it more sensitive to an electromagnetic field. Hence a weaker field, as it happens when one heaths the magnetic surface, can be used to record (and read) a bit.

It is good to see that as soon as scientists seem to have reached a barrier some other scientists find ways to circumvent the limitation.

As Lloyd has made clear in his paper on the “Ultimate physical limits of computation” that the physical limits to storage are very far away, some 400 years away assuming an evolution progressing at the same pace of the last fifty years.

Well, of course, the first thing coming to mind is that it is unlikely that you get to be blamed if the storage can no longer be read (or if you do you likely don’t care!).

However, our world is becoming more and more described through bits and with the present storage media they are unlikely to survive even 100 years. Compare this to the masterpieces from painters or writers. Can you imagine a world without a Van Gogh or Shakespeare heritage? Indeed, if Van Gogh and Shakespeare have chosen (not that they could) to store their masterpieces in jpg or doc files we won’t be able to enjoy them today.

Comparison of DNA vs MagTape cost taking into account the cost of writing with DNA and the time of content life

Scientists have been working on this issue, that is becoming ever more pressing since today we have reached a digital store in the ZB figure. What can be used to preserve digital information forever (or at least 10,000 years)?

Just knock on Nature’s door.

Researchers at the EMBL-European BioInformatics Institute have succeeded in finding an effective way to write DNA code to store information.

The DNA is a very robust way to store information, we can read the DNA of animals that got extinct long ago, 10s and 100s thousands year ago.

With all the advances we had in the last decade reading DNA sequences has become easy and cheap. What has remained very difficult is to write specific sequences of DNA since the number of errors during the writing procedures are too high.

The researchers at the EMBL-EBI have succeeded in developing a technique that is much more robust so that even 4 consecutive errors can be intercepted and recovered.

Do not expect, however, to get rid of your hard disc anytime soon to replace it with a DNA cup, even though that cup would be able to store over 100 million hours of hd video.

As it is shown in the graph, the cost is still very high, and it needs to decrease at least a hundred folds to become competitive with a magnetic tape. But of course, cost reduction is what we have become used to in ICT!

Look at the smoothness of the right hand side, comparing it to the left hand side

Researchers at the Singapore University have found a way to create self assembling surfaces that can be used as a substrate for storage.

There are two interesting advantages offered by this approach. First it can lead to a storage density of 10 Tbits per square inch and secondly, and to me even more interesting, it would allow to lay storage support on many kind of surfaces thus enabling “memory” on any object.

Self assembling is a well known technique but the challenge is to have it happening in a predictable way and in large volume.

Researchers at the Singapore University found out that in case of developing a uniform patterned surface like the one that can be used for storage the key issue is to have a really smooth surface onto which the particles can self assemble. It turns out that the surface needs to be smooth within a 5 Angstrom tolerance, that is a 5 atoms tolerance. What they discover is how to create this kind of smoothness…

This is also showing the level of progress scientists have made in their ability to control materials at atomic level.

Transparent flexible memory

Researchers at the RICE university have managed to create a transparent memory made by silicon oxide sandwiched between graphene electrodes and supported by plastic. The resulting product, as you can see in the photo on the left, is completely transparent.

The work is the result of a 5 year program that created all the tiny pieces needed for such a memory.

The researched started to find a way to store bits in a material that was impervious to cosmic radiations and a few chips have indeed been tested aboard the Space Station in orbit.

The researches noted that a strong voltage could break strings of carbon atoms but a lower voltage would cause them to break and then to rejoin. Associating 1 to interruption and 0 to continuity can make for a storage device.

Clearly, moving from a conceptual possibility to something that can work, and be produces by industry at low cost and in large amount, is quite a different matter. This explains the  time it took the researchers to reach this point.

A bonus, not originally planned, is the transparency. Silicon oxide is like glass, it is transparent and the carbon layer is just one atom thick, and so it is transparent as well.

A transparent memory can be placed in windshields keeping them transparent. Today we have means to project on the windshield information so that one can overlay information on the line of sight of the driver.

This overlaid projection, however, is no particularly good and requires a lot of power to make the information visible in daylight. It would be much better if one can have the windshield doubling up as a screen.

Indeed, this is what I saw at a both at the ITU workshop in Dubai. I took a picture of it and it is shown here. The visibility is very good also in full daylight so it can really be a solution in a few years.

I have attended a meeting organized by SAP on Big Data and one of the speaker noticed that the cost of storage in flash memory used to be 46 euro in 1990. That cost went down consistently and was less than 1c of euro in 2010. Over twenty years it decreased 4,600 times. If you apply that scaling to the tour Eiffel (the meeting was in Paris….) you would have now a tour that can be used as a keychain….

By comparison, the speed to access data on hard disks has not decreased at all! It was 25 ms in 1973 (the first magnetic hard disk produced by IBM) and it is now around 8 ms: 3 times less.

This is leading to consider disposing magnetic hard disks for processing data, and to shift to flash memory whose access time can be measured in ns (a million time faster!). SInce the cost of flash memory keeps decreasing it makes sense to move in this direction.

It is interesting to notice that as this shift occurs and the Cloud is also becoming more and more the “storage” there will be an increasing demand on faster connection: to get a nanosecond access time you would need a 1Gbps velocity. Notice how this is much more than 1 Gbps bandwidth since it includes overhead created by the IP protocol and, most important in the case of small transactions, the delay in the access (latency).

For this latter new architectures are needed, with full optical path.
Of course this does not matter to most of us but it may matter in some specific biz application. Even with the best architecture we can imagine today, however, the adding up of latency, protocol overhead and access drivers leads to a factor of at least 10,000 times slower data retrieval when using the Cloud in comparison with local flash memory and we are talking about the best of the best. If we look at the average it is probably over 10 million times slower than flash and 2 times slower than using local hard disks on a single byte access. Clearly, these differences shrink once we access MB since most of the delays are only seen at the start of the access and for all the other transfer only the transmission speed matters.

It is now several years that scientists have wondered if DNA can be used as a storing medium. It has been used by Nature to store the “code of life”, the billions of instruction directing the development and up-keeping of an individual, be it a rose, a fly or a human being.

Density of data storage in different medium

Several experiments, some of them already reported in this blog, have shown the feasibility and some practical applications already exists (like the watermarking of agricultural products to identify their origin).

However, this is probably the first time that researchers have managed to store an entire book, text and images, on a DNA strand.

The feat has been accomplished by Harvard researchers at the Wyss Institute for Biomedical Engineering.

Geneticist George Church has written a book, Regenesis: How Synthetic Biology Will Reinvent Nature and Ourselves, that will hit the shelves on October 2nd, and he and his team have coded the book also on a DNA strand. At a certain point they have also considered attaching the  DNA string to the paper book but then they decided to not do that because of still lingering concern on the proper use of DNA engineering techniques.

Interestingly, once the book has been copied in a DNA form, that DNA string has been duplicated, through usual DNA replicators in billions of copies, making this book the one having most copies ever!

It has been estimated that 4 grams of DNA would be sufficient to store the whole information produced in a year by humankind and that storage could resist thousands of years with no degradation whatsoever, making it both the cheapest and the most durable media ever used.

The drawback, with respect to other storage media, is the length of time it takes to store and read the information, a time that has to be measured in hours today, and in minutes by the end of this decade, versus the few microseconds it takes when using a flash memory or an optical/magnetic storage media.

Nevertheless, although DNA storage may never be a good solution for storing a movie or a song, it may be a good one for recording massive amount of information that need to be preserved over thousands of years.

You may want to look at the video clip telling the story in the researchers words;

[vimeo 47615970]