Defense Case Study
Liquidware developed and implemented a robust, modular computing technology for use in military applications. Dubbed “the bulletproof motherboard”, the system ensures continued function even after sustaining significant damage.
An academic research team received funding from Department of Defense and Department of Energy to design and develop an autonomous, massively parallel, and easily scalable computing system. The concept was simple: create a physical network of miniaturized, bare-bones embedded motherboards, which would each be able to autonomously send code, synchronize processing, and re-route resources should a node in the network fail. By building redundancy into the hardware, processing power and capacity are not limited by a single point of failure, as many standalone computing systems are.
The theory revolved around reframing the traditional von Neumann computational architecture. Traditionally, a computer comprises a central processing unit, data storage, and discrete random access memory. When any of the modules fail, the entire system is rendered non-functional. The alternate approach is to build a system from numerous ‘microcomputers’, each containing a small portion of processing power, data storage, and memory.
Translating academic research concepts into physical hardware constructs and software algorithms required a solid grasp of the underlying theory as well as systems architecture expertise. Liquidware worked with the academic team to develop a basic, high-connectivity prototype for proof-of-concept and benchmark testing.
In this role, Liquidware partnered with a cutting-edge research team to convert academic findings into innovative solutions. Translating a theoretical, scientific concept into physical devices with practical applications called for a hardware design team well-versed in computational architecture and hardware algorithm optimization.
Given the conceptual and experimental nature of the project, Liquidware applied computational architecture expertise to gain a better understanding of the problem at hand. This involved working closely with the academic team to create a theoretical blueprint of hardware behaviors. In particular, each modular motherboard would have extensive, orientation-sensitive connectors, allowing rapid transmission of board status, identity, as well as programming.
An integral aspect of creating a modular motherboard that would be viable in the field is small form factor. Consequently, Liquidware mapped out design requirements at a larger scale, ensured functionality, and then proceeded to execute a form factor and aesthetic optimization phase, bringing the final size of a 72 MHz single board computer down to less than 4 square inches. These boards could be linked in “grids” via 7-pin headers, allowing for power and data transmission, and routing processing power and data output where spare CPU cycles were available.
Achieving scale production on such a complex and miniaturized board design required close supply chain and vendor management. The production phase moved forward in parallel with a joint algorithmic exploration phase. Liquidware hosted multiple coding workshops with the academic team focused on developing benchmark tests and a parallel coding paradigm to ensure maximal performance, both for field use as well as future funding demonstrations.
The prototypes were deployed in experimental scenarios, both for further software development, as well as for use ruggedized environmental evaluation. Based on initial feedback, different versions of the board were customized for a variety of use cases. Success in multiple test cases also drew interest from other branches of the military, as well as private sector firms interested in deploying “smart-routing” redundant computer networks:
- In response to decreased power to the grid, it can turn off precisely the right number of nodes to stabilize itself, and continue functioning at lower power
- If a single node/unit is shorted suddenly, its neighbors will sever their own outbound power before the fault propagates and takes down the system
- If a single node/unit is destroyed within the grid system, its neighbors will adaptively re-route power and logic control to continue functioning
Following hardware development completion, Liquidware worked closely with research teams to build out the software coding platform required to take advantage of parallel processing algorithms.
Liquidware has received additional funding to develop a second-generation modular motherboard, retaining the small, modular form factor, while boosting processing power, memory, data storage, and peripheral availability on a single board.