Ask not what you can do for your Cloud, but what your Cloud can do for you!
Mayday, Mayday, Cloud under attack!

Hollywood & NASA Paving the Way: The Rise of the Billion-Node Machine Part 2

Imported from published August 1 2010

In a previous blog 'Planetary Cloud: The Rise of the Billion-Node Machine', I described a possible evolution of Public Clouds into Planetary IT complexes, where billions of nodes are available planet-wide to work on some of the most challenging problems of our times – what I affectionately called the Billion-Node Machine.

For those of you that think this is far in the future or just plain nonsense, think again! While the private sector is working on using Cloud technologies to take advantage of scale economies and effective working models to underpin IT strategy, there are some institutions on this planet already taking the leap forward into these billion-node futures.

NASA, a well-known pioneer around the world, has been collaborating with Microsoft to bring us the Worldwide Telescope project. The aim was to bring ‘Mars to our planet’ in the form of high resolution imagery processed through Nebula, the NASA cloud. There is an interesting video of the project background containing interviews with the engineers and designers.

Such projects are difficult to accomplish with ‘normal’ scaled Clouds - we are talking here about processing hundreds of Terabytes in near real-time, and that is just the tip of the iceberg. If that imagery needs to be rendered in different formats for all devices / formats available on the Internet then this could be many times the original imagery size from the Mars mission beamed to Earth through space. NASA says there are '15,000 tera-pixel photos that are then mosaic-ed into a half billion PNG images via the Nebula cloud'.

Nebula is interesting in that the needs of NASA are so extreme, extreme solutions are required to solve the problems that are quite literally ‘out of this world’. Having started out trying to copy the Amazon EC2 Cloud infrastructure; finding it too limiting to accommodate their innovation needs and lacking the unrestrained free-hand to change things as needed; the Ubuntu Enterprise Cloud technology was then used. However, this was also not open or scalable enough to support the NASA extreme needs. NASA is now working with Rackspace and others to create the OpenStack platform.

Just that we understand the desires that NASA has in mind to get a sense of scale. NASA is working to build an infrastructure cloud spanning over ‘1 million physical machines and 60 million virtual servers’. That is a far cry from the thousands of virtual machines organizations are currently running. Even the Google infrastructure is small against that backdrop, at least for the moment.

This convergence of HPC (High Performance Computing – Grid) worlds, where systems with a single system image of over 10,000 cores work simultaneously on a problem, and the Public Cloud as we know it for typical computing/storage workloads seen in companies together with consumer demands is creating a new breed of Cloud – the Billion-Node Machine starts to take shape!

To see how that is being used in real life, one needs to look outside of typical corporate boardrooms. Some famous names apart from NASA are already taking advantage of such elastic and massively scalable environments. DreamWorks, the creator of such wonders as Shrek and Madagascar, is taking advantage of supercomputer class resources in the New Mexico Applications Centre (NMCAC). This may supersede its own impressive infrastructure farms consisting of thousands of servers.

Key enabling infrastructure such as LambdaRail optical network and the supercomputer Encanto, based on more than 3,500 Intel quad-core processors, 28TB of memory and 172TB of storage. That is over 14,000 cores able to work simultaneously on a problem – or sliced and diced as needed in real-time to host virtual machines. This is not the largest or most powerful supercomputer resource. Nonetheless it costs around 2 million dollars annually to power and cool this system alone. Well beyond the reach of most mere mortals.

Just to put those studio resources into a frame of reference, that everyone has access to, take the creation of some of the animations produced over the last half decade. Computers used in the development of the 2006 Pixar Cars film were four times faster than those used in The Incredibles (2004) and 1,000 times faster than those used in Toy Story (1995). To build the cars in the movie, the animators utilized compute platforms that one would use in the design of real-world automobiles - convergence!

Avatar, the pre-cursor of the 3D film wave we are currently experiencing, was produced in many different formats for viewing in 2D/3D, and literally translated human movement into a realistic digital experience (instead of trying to copy human movement). This film required extensive computer tooling; the exact mix is of course secret ;-) Such films were simply impossible to create even some years ago.

The billion-node machine allows other ideas to see the light of day. It is an innovation catalyst, a tool to allow people to flex their minds on scales previously impossible (not unimaginable though). A chance to bring computing resources to every point on the globe, and have consumers/users literally only needing a connection and a display/input tool – even the need for powerful local CPUs is bypassed.

If the dreams, that are films, are being created, rendered and produced in formats on the fly in such a way, then perhaps the future will belong to advanced digital technologies presented in life-like holograms all powered by a billion-nodes with real-time movements.

Perhaps with such resources, each household will have its own virtual datacenter that can grow/shrink as needed. Add to that advances in artificial intelligence and voice recognition and perhaps multiple holographic images can be created able to interact in real time, with memos dictated in real-time bypassing crude email.

This would be possible to use as full featured telepresence system by projecting millions of people around the world into meetings without them needing to fly/drive physically to the locations. That is a far cry from the crude telepresence systems we currently have. If those holograms could also move physically around buildings and the imagery sent back to the person being projected then that is as good as being there (Emperor Palpatine as a hologram moving on a walking robot in Star Wars: Phantom Menace - was good enough for him).

Perhaps that is the 'killer app' for the billion-node machine. The ability to be able to code/develop/share ideas literally face-2-face is an amazing competitive advantage, well beyond simple outsourcing. Perhaps the costs of using local talent are not as high as currently thought – outsourcing may disappear as a term in general use. We don’t have the matter transporter yet, but this is pretty close to that experience!

Many have seen the holodecks in Star Trek – Next Generation/Voyager (with Data enjoying a poker game with Albert Einstein, Sir Isaac Newton and Steven Hawking), or the amazing utilization in Minority Report (2002) that had a live demo of some of the technologies in 2010!. This is really cool and great to understand the critical role of user interfaces!

I used the word 'perhaps' very often as that is the space where innovation and technology cusps meet. Are these ideas really so far-fetched when that level of computing power starts to become available? The disruptive market effects are already being felt with NASA undertaking support of private sector joint initiatives with the Worldwide Telescope – looking 'out there' as it were.

What will happen when this machine is turned to looking inside – the atomic level of reality? What will happen in the gene technology worlds, and its complementary effects with nanotechnology? Could such technology have helped in cleanup efforts from the BP oil leak in the Gulf of Mexico?

The creation and adoption of billion-node constructs and planetary IT specifically need to be speeded up. The lack of processing power should no longer be an innovation inhibitor - and some problems require everyone's help. Let's get some global economies of scale in here.

EMC is doing its bit by stimulating discussions and idea submission of the practical uses of such constructs or their underlying technologies - what we call Innovation Showcase. It doesn't matter how wild they are or where they come from! The whole IT industry is in a frenzy of innovation where the vision of the Cloud is being translated into innovative and practical solutions across industries and borders - Global Innovation!

Things are starting to get seriously interesting. It is well worth watching what the pioneers, the early-adopters at this planetary scale, are doing.

It provides insight into how they are tackling the processing and handling of petabytes of information in near real time, the distribution of visualization of such volumes of data to extract information, and how their actions are affecting the wider Cloud technologies markets and vendor maneuvering.

Perhaps when space agencies start to cooperate at the level of their IT processing capabilities, then we may see the first billion-node machine emerge! The race is officially on!


The opinions expressed here are my personal opinions. Content published here is not read or approved in advance by EMC and does not necessarily reflect the views and opinions of EMC.


Feed You can follow this conversation by subscribing to the comment feed for this post.

The comments to this entry are closed.