AXIOM EXECUTIVE SUMMARY
We are entering the Cyber-Physical age, in which both objects and people will become nodes of the same digital network for exchanging information. Therefore, the general expectation is that “things” or systems will become somewhat smart as people, allowing a rapid and close interaction not only system-system, but also human-system, system-human. Moreover, through smart systems, the human behavior is improved and simplified. More scientifically, we expect that such Cyber-Physical Systems (CPS) will at least react in real-time, provide enough computational power for the assigned tasks, consume the least possible energy for such task (energy efficiency), scale up through modularity, allow for an easy programmability across performance scaling, and exploit at best existing standards at minimal costs. The whole set of these expectations impose scientific and technological challenges that need to be properly addressed.
The AXIOM project (Agile, eXtensible, fast I/O Module) aims at researching new software/hardware architectures for CPSs to meet the above expectations.
The current solutions for providing enough computational power are mainly based on multi- or many-core architectures. For example, some current research projects (such as ADEPT or P-SOCRATES) are already investigating how to join efforts from the High-Performance Computing (HPC) and the Embedded Computing domains, which are both focused on high power efficiency, while GPUs and new Dataflow platforms such as Maxeler, or in general FPGAs, are claimed as the most energy efficient ones. For this reason, our research will not be limited to only one kind of technology, but it will start from power efficient multi-cores such as ARM cores and FPGA accelerators on the same die as in the Xilinx Zynq.
Programmability: we observe that current toolchains for CPSs are too specific to the selected domain, involving customized or expensive Software Development Kits (SDKs). Recent examples such as Raspberry Pi or UDOO (the latter being funded by two of AXIOM partners) represent successful examples of eliminating the tedious learning steps of customized SDKs and their associated high costs, relying on totally open hardware designs and open-source software. This has the potential to open the development to a broader community not confined to experts but also open to other end-users like the wider educational sector and the “Makers” community.
Parallelism: several solutions have been proposed during the last decades. However, a unanimous consensus on the best solution has not been reached. In this project, we build on the consolidated expertise of partner BSC in the HPC domain to bring, also in the CPSs domain, the simplicity and scalability of the OpenMP framework. In particular, we will focus on its extension to the StarSs paradigm, called OmpSs.
Interaction Design: one additional key element of this proposal is the availability of expertise from partner UNISI on Interaction Design. This expertise will ensure a high usability of our CPS and the adoption of appropriate patterns (also from a psychological point of view) for the interaction and the communication with the humans.
Computer Architecture: previous experimentation brought the UDOO as the first world-wide board integrating a powerful quad-core ARM (able to run both Linux and Android smoothly) and the Arduino Due compatible interface, for an easy plug-and-play of a huge number of sensors and actuators. Our project will push further the idea of a general, modular, reusable single module capable of providing further specialization through the reconfigurability features of the FPGAs.
Interconnects: not only our CPS will include classical connectivity (e.g., Internet), but we will bring modularity to the next level by allowing to build more compute intensive systems through a low-cost scalable high-speed interconnect. This interconnect, subject of research and design during the project, will be developed by partner FORTH, on the experience coming from the ENCORE project and used in the Formic platform for creating a 512-core system. Such connectivity will allow to build (or upgrade at a later moment) flexible and low-cost systems with simplicity by re-using the same basic (small) module without the need of costly connectors and cables.
Real-Time: another relevant aspect is the necessity of properly managing actions in real-time and through an Operating System (such as Linux). One key partner (EVIDENCE) will provide the consolidated expertise to integrate real-time management having recently succeeded in integrating the Earliest Deadline First (EDF) real-time scheduling algorithm in Linux 3.14. Moreover, EVIDENCE will integrate features in the OS drivers to allow a fast balancing of the work across the basic modules through the high speed interconnects. Current solutions for work balancing across distributed systems are expensive, too customized, or too difficult to program (with paradigms such as MPI). We want to investigate the usage of OmpSs on top of a variant of a Distributed Shared Memory (DSM) that will be topic of research and development during the project.
From an exploitation point of view, one partner (SECO) has the full capability to build and commercialize the module designed in the project. Being member of the European standardization effort SGET, SECO will be capable of contributing to a European standard for the module.
Moreover, partner UNISI will provide evaluation and design space exploration technologies consolidated in recent experience in the TERAFLUX and ERA projects and further develop open-source components for the thread scheduling and consistency. UNISI will provide at any time during the project the evaluations and measurements that will drive the development of the software parts and the selection of the most appropriate hardware architecture and interconnect.
One of our partners (BSC) will further extend the toolchain and work in close collaboration with industrial partners for developing the use cases to demonstrate the effectiveness of our targeted CPS. In particular, we wish to demonstrate a scenario involving Smart Video-Surveillance and another one involving Smart-Living/Home.
For the Video-Surveillance, we may have to process a large amount of data (Video-Surveillance is also called “the Biggest Data” in that respect) for smart multi-focus recognition or coordination of multiple cameras towards a single event with context shift. The interested partner is HERTA.
For the Smart-Living/Home, instead of envisioning fancy Domotic scenarios unaffordable by the largest part of the population, we will focus on a simpler system that could replace the omni-present ambient thermostat with a much smarter device (similar to the Nest concept). This device will be scalable from the small house to big buildings and may be able of acting as a small network server or proxy for a large number of existing or near future services in collaboration with the municipality or even in a peer-to-peer scenario. The interested partner in this case is VIMAR.
Partner UNISI will be in charge of coordinating the project, also due to its experience in coordinating large European projects such as TERAFLUX. BSC will lead the software toolchain development. EVIDENCE will lead the Runtime and Operating System development. SECO will lead and realize the AXIOM hardware, with the help of FORTH that will lead the interconnect design. The industrial partners are: EVIDENCE, HERTA, SECO, and VIMAR.