Periodic Reporting for period 3 - BRAINE (Big data pRocessing and Artificial Intelligence at the Network Edge)
Reporting period: 2022-05-01 to 2023-11-30
The BRAINE project’s overall aim is to boost the development of the Edge framework and, specifically, energy efficient hardware and AI empowered software systems, capable of processing Big Data at the Edge, supporting security, data privacy and sovereignty. BRAINE’s overall aim will be reached by targeting five fine-grained goals:
- Devising an EC infrastructure that offers control, computing, acceleration, storage, and 5G networking at the Edge and excels in scalability, agility, security, data privacy, and data sovereignty in Big Data and AI for low latency and mission-critical applications.
- Developing a future-proof Edge security framework and associated infrastructure. based on the latest software and hardware security technologies.
- Developing a distributed and partly-autonomous system that takes data privacy and sovereignty into account on each and every decision regarding workload placement, data transfer, and computation, while guaranteeing interoperability with the environment.
- Developing a heterogeneous, energy efficient Edge MicroDataCenter, suitable for stationed, mobile, and embedded autonomous applications, that goes beyond the current hardware and software architectures and offers Big Data processing and AI capabilities at the Edge.
- Testing and demonstrating the effectiveness and generality of the BRAINE approach by evaluating multiple real-world use cases and scenarios that exhibit the required scalability, security, efficiency, agility, and flexibility concerns.
Through the achievement of its goals, the BRAINE project will help to position Europe at the forefront of the intelligent edge computing field, enabling growth across many sectors (manufacturing, smart healthcare, and others). By lowering the barriers for utilising edge computing for artificial intelligence applications, BRAINE will open the door for European industries and SMEs to leverage state of the art technologies, driving their development and growth as industry leaders in their sectors.
- Devising an EC infrastructure that offers control, computing, acceleration, storage, and 5G networking at the Edge and excels in scalability, agility, security, data privacy, and data sovereignty in Big Data and AI for low latency and mission-critical applications.
- Developing a future-proof Edge security framework and associated infrastructure. based on the latest software and hardware security technologies.
- Developing a distributed and partly-autonomous system that takes data privacy and sovereignty into account on each and every decision regarding workload placement, data transfer, and computation, while guaranteeing interoperability with the environment.
- Developing a heterogeneous, energy efficient Edge MicroDataCenter, suitable for stationed, mobile, and embedded autonomous applications, that goes beyond the current hardware and software architectures and offers Big Data processing and AI capabilities at the Edge.
- Testing and demonstrating the effectiveness and generality of the BRAINE approach by evaluating multiple real-world use cases and scenarios that exhibit the required scalability, security, efficiency, agility, and flexibility concerns.
Through the achievement of its goals, the BRAINE project will help to position Europe at the forefront of the intelligent edge computing field, enabling growth across many sectors (manufacturing, smart healthcare, and others). By lowering the barriers for utilising edge computing for artificial intelligence applications, BRAINE will open the door for European industries and SMEs to leverage state of the art technologies, driving their development and growth as industry leaders in their sectors.
The BRAINE Project designed, developed, and validated an innovative edge computing hardware prototype named Edge Micro Data Center (EMDC). The BRAINE EMDC includes: innovative cooling systems, new boards containing the composable building blocks (i.e. GPU, ASIC, FPGA, CPU, memory), specifically designed embedded software solutions (i.e. firmware), and telemetry-based functions enabling monitoring capabilities and network awareness at the orchestration level. At the hardware and embedded software level, the BRAINE Project has also developed innovative crypto-accelerators for storage and secure communications, securitization of the communications between the EMDC and the rest of the network, integration of the EMDC with complementary systems for communication such as 5G.
At the software level, BRAINE designed, implemented, and validated an innovative components enabling (1) edge resource collaboration, (2) resource management and service deployment, (3) AI/ML-based workload placement, (4) efficient monitoring infrastructure, (5) secure data management framework for AI at the Edge, (5) efficiently handling data at the edge at scale, (6) workflow definition language and authoring tool, and (8) execution measurement and monitoring.
These software components have been successfully integrated in a comprehensive BRAINE software platform. A user guide for the final public release of the platform has also been released.
The overall BRAINE solution has been successfully validated against four use cases: healthcare assisted living (case 1), hyperconnected smart city (case 2), robotics in Factory 4.0 (case 3) and supply chain Industry 4.0 (case 4).
Concrete per-partner and joint exploitation and dissemination results have been achieved, successfully contributing to position Europe at the forefront of the intelligent edge computing field.
At the software level, BRAINE designed, implemented, and validated an innovative components enabling (1) edge resource collaboration, (2) resource management and service deployment, (3) AI/ML-based workload placement, (4) efficient monitoring infrastructure, (5) secure data management framework for AI at the Edge, (5) efficiently handling data at the edge at scale, (6) workflow definition language and authoring tool, and (8) execution measurement and monitoring.
These software components have been successfully integrated in a comprehensive BRAINE software platform. A user guide for the final public release of the platform has also been released.
The overall BRAINE solution has been successfully validated against four use cases: healthcare assisted living (case 1), hyperconnected smart city (case 2), robotics in Factory 4.0 (case 3) and supply chain Industry 4.0 (case 4).
Concrete per-partner and joint exploitation and dissemination results have been achieved, successfully contributing to position Europe at the forefront of the intelligent edge computing field.
· Definition, design, implementation, and validation of the innovative BRAINE EMDC hardware solution based on: (1) specifically designed monoblock chassis with innovative passive (i.e. fan-less) two-phase cooling system based on Loop Thermosyphon (LTS) and nanoparticles, lowering OPEX 20-30%; (2) new backplane/connector including both dual x8 lane Gen4 PCIe interface and 4 x 10/25G Ethernet link for unprecedented intra- and inter-EMDC throughput; (3) new compact card design fitting the proposed BRAINE common form factor and drastically reduce non-processing hardware redundancies.
Design, development, and successful validation of:
- EMDC Embedded Software based on open components (e.g. SONiC operating system for the embedded Ethernet switch) and efficient EMDC management infrastructure compliant with OCP (Open Compute Project).
· hardware accelerated innovative solutions for packet filtering, FPGA-based MAC layer, digital IP blocks of remote radio unit for accelerating 5G networking, and non-volatile memory express acceleration.
· inter-EMDC Communication infrastructure based on: (1) telemetry-based SDN-controlled solution relying on open disaggregated multi-vendor architecture; (2) innovative cost-effective filterless solution exploiting bi-directional transmission over a single fiber; (3) 5G software defined RAN (SD-RAN) with external radio unit (RU) for the different 3GPP functional splits, including split 6, 7.2-x.
· effective solution to develop services that enable applications to execute workloads utilizing Edge resources. Services include virtualization, automated deployment, smart workload placement, running and managing workloads, and monitoring of the resources and applications.
· smart, agile workload placement and orchestration that considers the unique characteristics of the Edge nodes, ensuring that applications can run reliably on the infrastructure and that the resources are used in an efficient way.
· distributed learning techniques preserving privacy and improving security on the Edge while balancing with computational and networking heterogeneous resource constraints.
· innovative in-network firewalling implementations
· CPU offloading of IPSec and transport security layer (TLS) protocols in smartNICs.
· Evaluation of energy efficiency of QKD considering the post-quantum security that is under benchmarking of the standardization bodies.
· Design of cryptographic acceleration for in fibre-based connectivity links for edge nodes. A link-security testbed including the EMDC module and the smart-NIC from Mellanox has been setup leveraging on a ring of dark fibres with quantum key distribution (QKD).
Overall, BRAINE provided a new vision for utilizing edge resources by enabling novel network-edge workload distribution schemes, predicting resource availability and workload demand, identifying trends, and taking proactive actions are all aspects of the novel workload distribution. Through BRAINE, edge computing can reduce this projected energy consumption by offloading many of the AI functions next to the end-users.
Design, development, and successful validation of:
- EMDC Embedded Software based on open components (e.g. SONiC operating system for the embedded Ethernet switch) and efficient EMDC management infrastructure compliant with OCP (Open Compute Project).
· hardware accelerated innovative solutions for packet filtering, FPGA-based MAC layer, digital IP blocks of remote radio unit for accelerating 5G networking, and non-volatile memory express acceleration.
· inter-EMDC Communication infrastructure based on: (1) telemetry-based SDN-controlled solution relying on open disaggregated multi-vendor architecture; (2) innovative cost-effective filterless solution exploiting bi-directional transmission over a single fiber; (3) 5G software defined RAN (SD-RAN) with external radio unit (RU) for the different 3GPP functional splits, including split 6, 7.2-x.
· effective solution to develop services that enable applications to execute workloads utilizing Edge resources. Services include virtualization, automated deployment, smart workload placement, running and managing workloads, and monitoring of the resources and applications.
· smart, agile workload placement and orchestration that considers the unique characteristics of the Edge nodes, ensuring that applications can run reliably on the infrastructure and that the resources are used in an efficient way.
· distributed learning techniques preserving privacy and improving security on the Edge while balancing with computational and networking heterogeneous resource constraints.
· innovative in-network firewalling implementations
· CPU offloading of IPSec and transport security layer (TLS) protocols in smartNICs.
· Evaluation of energy efficiency of QKD considering the post-quantum security that is under benchmarking of the standardization bodies.
· Design of cryptographic acceleration for in fibre-based connectivity links for edge nodes. A link-security testbed including the EMDC module and the smart-NIC from Mellanox has been setup leveraging on a ring of dark fibres with quantum key distribution (QKD).
Overall, BRAINE provided a new vision for utilizing edge resources by enabling novel network-edge workload distribution schemes, predicting resource availability and workload demand, identifying trends, and taking proactive actions are all aspects of the novel workload distribution. Through BRAINE, edge computing can reduce this projected energy consumption by offloading many of the AI functions next to the end-users.