Software robotics and drones – from design to concept
During the rise of technology trends like software-defined infrastructure and hardware disaggregation in 2015, I started looking into what could be the next step in the evolution of datacenter hardware, trying to envision where we would be around 2020.
Given that the amount of data we process is growing exponentially, it is said that by 2020 the internet could be 8-16 times the size of what it was 2015. As we are approaching the age of 5G and Internet of things (IoT), the question is, what if Moore’s law is not keeping up with the speed of data growth? Will we end up drowning in our own data with datacenters all over the place?
There is evidence that some datacenter facilities have grown to the size of small cities like Langfang in China with its 6,3 million Square feet, and it just keeps growing.
Are we getting to the point where existing systems are reaching their effective limits?
So, let’s say that we see an increase in the need of datacenter facilities, and I mean different types of facilities, from bigger centralized sites to smaller distributed edge sites that are closer to the end user. This would require that the equipment would be more efficient in terms of deployment, scaling and cost. To solve this problem, we may require a completely different type of infrastructure.
Early sketches of the disaggregated tile idea (2015)
Boxes in racks with spaghetti in the back
I continuously keep asking why things look the way they do, so what about these racks with boxes?
When it comes to the 19-inch rack that we see in all large server setups today, it is a format that has been around since the inception of server systems, and in my opinion it feels very dated. In fact, it dates all the way back to the beginning of the last century, established by AT&T and was originally used for equipment for toll cables.
My point is, much of what we see around us is tied to design decisions and compromises of the past.
The question we need to ask ourselves however, is if we were to design a server/datacenter and its infrastructure for the first time today based on what we know, would we end up with a steel frame rack that measures 19 inches in width, and no higher than an average human being can reach to the top of it, and then slot compute and storage sleds into it with spaghetti hanging from the back?
What could software-defined mean for the hardware?
When examining trend forecasts, a common thread in the discussions that I found was “software-defined datacenter”. But what nobody asked was what this could really mean for the hardware and the way we build our infrastructures.
I looked back at two of the most disruptive technologies at that time: optical interconnection and Software-Defined Infrastructure (SDI), which means for example, that the compute unit is optically interconnected with the hard drive units, and where resources can be pooled and connected with software. I then asked myself, how we could optimize and simplify and let the hardware reach its full potential based on these two technologies, and again, what does software-defined really mean for the hardware and the datacenter as we know it?
Tile concept illustration (2015)
What if we could envision a completely new format/platform that is generic, scalable, and simple enough to be used in any scenario – from hyperscale to edge/micro servers – to accommodate these challenges?
One way to do it could be having processors, memory, storage, network, etc. decoupled into modules/tiles that were fully optically interconnected down to chipset, and where servers could be virtually assembled from pools of data. This would basically, allow us to take the decoupling of components (disaggregation) to the next level!
This would mean that we would no longer be limited by the size of the motherboard and the rack, and any possible configuration could be virtually assembled based on need, from the most performance optimized super computer down to the simplest IoT application with the same components, like a server on demand…
The conceptual design proposal presented in early 2016 was based therefore on the idea of a disaggregation of components into 20x20cm tiles, stripped down to the minimum and optimized without any boxes around them, instead completely open with bare circuit boards with components. When entering such a datacenter/server it would be like stepping inside a huge computer with all the visible exposed components.
The idea of the disaggregated tile was that it would take a flexible format that could be used in any sort of configuration, from a traditional rack mount to a smaller edge compute enclosure.
Tile concept visualization (2015)
Before the summer 2016, the Industrial design department was challenged to come up with a conceptual new datacenter layout. A datacenter that was more of digital factory and less like a barn with chickens on the floor. A datacenter that was fully automated and absent of human beings, and a visionary concept demo of a future digital infrastructure to show at Mobile World Congress 2017.
We wanted to emphasize/explore the potential of some of the disruptive technologies like physical automation/robotics and disaggregation/SDI with design, and communicate these ideas through a speculative concept model. Basically, a sort of physical representation of a future vision of a datacenter and the digital infrastructure.
Early exploratory sketches of datacenter automation ideas (2016)
Rather than using design for problem solving, usability and sexy styling, which is a common understanding of what design is used for, I find it is interesting to use design as a tool to make up ideas and possible scenarios based on disruptive technologies, by challenging assumptions and to ask the “why” and “what if” questions.
This is a way to speculate about how things could eventually be and to imagine possible and desirable futures, instead of the common way of predicting and forecasting.
The “Office” (2016)
Why and what if?
Of course, we started asking ourselves a lot of questions:
What if we used some of these disruptive technologies and combined them, what could it be like?
Why does a datacenter look the way it does? Why only stack the servers to human height? Why cool the whole datacenter and not just around the hotspots? Why have humans at all in there?
How would the design look if we based it on the anatomy of a robot instead of a human?
We looked into different areas of the latest research on cooling, disaggregation, automation, optical interconnection, and more, and started composing a concept based on some of these ingredients.
The idea with the disaggregated tiles from 2015 turned out to fit very well into this.
Drone delivery concept illustration (2017)
Inspired by supercomputers, automated tape libraries and high-rise warehouses, we started sketching up ideas of different infrastructures and tile configurations as a base to further discuss possible future scenarios.
One thing that we wanted to emphasize with this concept was the scalability in height in combination with automation, just like one floor storage facilities operated by humans have developed to high-rise storage operated by robots.
That’s when the idea of the flat panel server wall with tiles came up among many other ideas.
The following months were some intense and exciting ones…
Cylinder concept illustration (2016)
Cylinder concept visualization (2017)
Illustration showing the optically interconnected software defined component tiles (2017)
It is important to mention that this concept was not an exacting proposal of how it should be. Rather the goal was to create a thought-provoking showcase of merged ideas of a theoretically possible future scenario. It was supposed to represent the conceptual thoughts in a way that it could easily be explained and understood on a physical level. A show piece that should question our common assumptions and invite discussions and debate.
Flat panel concept illustration (2016)
Early MWC installation proposal 10X8m high (2016)
The aesthetics of the installation were supposed to look engineered and organized in a performance optimized manner. However, in reality, the tiles would probably be distributed in a much more space efficient way, but for this case were designed to expose the components and to make the idea more self-explanatory.
The “Wall” test site (2017)
We should be proud to say we were first to make a physical demo/concept out of this kind of idea, and I believe that design can be an excellent tool to facilitate a dreaming process that unlocks people´s imaginations and minds, that can inspire and support the development of what we do.
The future of digital infrastructure design is still unclear, and only time will let us know what becomes reality. What we see today is probably only the very beginning of a long evolution until we reach the final destination of digital infrastructure, and we have yet to know if it will be fully automated. I suppose it is just a matter of time…
What is Ericsson doing with future digital infrastructure currently? Check out NFVi and SDI in action!