Ce topic appartient à l'appel HORIZON-JU-SNS-2022
Identifiant du topic: HORIZON-JU-SNS-2022-STREAM-B-01-01

System Architecture

Type d'action : HORIZON JU Research and Innovation Actions
Nombre d'étapes : Single stage
Date d'ouverture : 18 janvier 2022
Date de clôture : 26 avril 2022 17:00
Budget : €24 000 000
Call : HORIZON-JU-SNS-2022
Call Identifier : HORIZON-JU-SNS-2022
Description :

ExpectedOutcome:

The expected outcomes support the vision of a massively digitised economy and society calling for intelligent connectivity and service provision across a huge number of heterogeneous domains, resources, and with an unlimited number of application requirements. These cover the availability of:

  • An overall system functional architecture to cater for the expected extreme 6G use cases.
  • An architecture able to break the boundaries between different infrastructure, service, business, and application environments, capable for a unified service provision across heterogeneous communication and computing environments.
  • Solutions for inter-computing beyond the inter-networking capabilities of the Internet, making possible the execution of services across multiple heterogeneous but seamlessly inter-working domains, each possibly applying different policies (e.g., in terms of security, routing, access to resources, etc.), routing mechanisms, access mode to application services, etc.
  • An internet-like architecture supporting much higher dynamics and versatility for its topology and service instantiation while significantly lowering energy consumption.
  • Architecture and technologies enabling the connectivity and service infrastructure to be programmable with a single, unifying, and open controllability framework, spanning all resources a tenant is authorized to control, including resources from currently separate and heterogeneous domains, such as enterprise and telecom networks, virtual and physical, data centres and routers, satellites, and terrestrial nodes.

Objective:

Please refer to the "Specific Challenges and Objectives" section for Stream B in the Work Programme, available under ‘Topic Conditions and Documents - Additional Documents’.

Scope:

The scope covers the realisation of a unified and open communication and computing architecture beyond the current SBA capabilities. Such architecture will enable seamless operations and service execution across a multiplicity of heterogeneous domains, infrastructures, services, business, and application heterogeneous domains, whilst providing secure and reliable scalability towards an unlimited number of application requirements. Hence paving the way towards massive digitisation. It offers a consistent/reliable programmable environment enabling “tailor made” implementation of various tenants’ requirements. Focus is on complementary issues as follows:

  • Technologies for scaling Inter-computing systems: The work addresses technologies and system architectures able to efficiently span all resources, regardless of their type (compute, storage, networking), nature (virtual/physical) or location (remote/local/cross domains), dynamically adding and removing resources as they come and go. System control solution aspects effectively handling authentication, naming, addressing, routing and related functions for massive number of elements may also be examined. Resource control takes an end-to-end approach covering the device/IoT domains with service deployment at the deepest possible “edge” in the user vicinity. The work enables a fusion of data communication with distributed computing, transforming the best effort Internet into a sustainable, greener inter-computing system. The needs of service developers and end-users may be considered, such as time-to-market, continuous functional improvements, increasing volumes of data collection for AI, the overall energy cost of providing added intelligence, as well as software foundations and interfaces, with a design-to-cost and design-to-energy efficiency framework.
  • Control and controllability separation: Infrastructure programmability is creating a new level of decoupling between the platform delivering the service and the service elements. Network virtualization brings additional degrees of freedom in flow processing and combines edge and network in one logical entity. It addresses optimal system programmability deployment on top of a shared, distributed multi-stakeholder infrastructure (at a horizontal and vertical levels). This is composed of different resources, shared by instantiation from other executed services or slices, while establishing system integrity and self-preservation in runtime for a distributed, dynamic resource environment. The architecture should be able to manage elements of different configurations and implementations, different active modules, deployed on top of more generically capable resource pool, in a multistakeholder environment. Data quality and trust levels are key parameters to consider as well as potential disruptions in network/service operation and how the overall architecture should respond to these to meet an appropriate resilience level for the service providers and the end users. Concretely, the aspect to be explored here is the separation[1] of service- and tenant-specific control in terms of the generic, autonomic, unified infrastructure and resource controllability featuring high resilience facing resource dynamics, and providing equivalents of so-called “protected modes” known from other execution environments.
  • Frictionless inter-domain resource management: The work addresses frictionless and optimised resource management and orchestration across multi-stakeholder cloud, edge, and fog platforms to meet specific application requirements (e.g., latency, performance and other relevant 6G KPI’s), with efficient utilisation of resources for both service provider and verticals operating under specific constrains (e.g., resilience, timing, cost, energy, CO2, etc). It includes capability to swap computation loads as a function of efficiency and local energy availability under strict resilience and timing requirements. To address resource control, including resource pooling, service request scheduling and conflict resolution, AI is expected to play a relevant role (e.g., network orchestration).
  • Native integration of AI for telecommunications. The massive adoption of AI tools will exacerbate the problem of energy consumption of the ICT infrastructure. Native integration of AI/ML is in scope to implement adaptive decision making at different time scales with expected impact on energy and performance efficiency gains for such distributed multi-stakeholders’ systems. The adoption of these tools may trigger changes in the existing architectures. Therefore, it will be crucial to devise energy efficient architectures and computation algorithms to have energetically sustainable communication and computing paradigms for future mobile networks that adequately explore artificial intelligence technologies. Research needs to be done on: i) distributed edge AI solutions, covering consensus convergence, resource limitations, localized data management, transfer learning; ii) adequate development of training data for telecommunications; iii) AI security and comprehensibility of ML for the applications identified above; iv) strongly distributed AI/ML instrumentation integrated at the architecture layer.
  • New Data Transfer Paradigms with deep Edge integration: The work considers systems where edge, access and cloud are increasingly undistinguishable (i.e., used homogeneously by the service layer). This work addresses edge-specific requirements originating for example from IoT devices that are service dedicated, intelligent and are yet resource constraint (e.g., micro-electronics for battery driven components). This yields a richness of resources that are challenging when being integrated into a common resource worldview. To support this, novel suitable switching, resource management and scheduling mechanisms that take all constraints into account, whilst relying on edge-specific control agents enabling the enforcement of the policies underlying the switching decisions and scheduling solutions. This further includes new IoT device management techniques that are adapted to the evolving distributed architectures for IoT systems based on an open device management ecosystem. The complexity originating from the diversity of different types of physical devices and communication links is also considered (sensors, user terminals, edge and data centre computers, optical and electrical switches, fibre, and satellite links) together with the diversity of service requirements. Highly scalable resource management techniques with deep compute/communication integration across multiple domain components are needed.
  • Improve data plane performance: The work addresses innovative protocols in view of overcoming known Internet limitations as originating from new scenarios and vertical requirements (ultra-low latency, extreme mobility, ultra-high data rates, integration of end-terminals, controlled security, space applications), It addresses functional improvements of the basic transport mechanisms with guaranteed packet delivery, increased dynamics in network topologies as well as compute resources and the resulting required flexibility in routing, while also considering security and precision delivery. These protocols should be able to flexibly operate in local/global architectures and provide primitives to perform the integration of new localized environments in an intelligent ICT infrastructure (either as overlay or underlay protocols). Brown or green slate approaches may be considered, as well as challenges from hardware architectures, protocol design, semantic approaches, optimized software focused data planes, as well as a clear migration strategy from legacy technologies.
  • Deterministic Networking: The work opens an entirely new class of innovative application by enabling deterministic networking, beyond today’s best efforts Internet characteristics. It goes well beyond the current work of 3GPP Release 16 targeting industrial applications. It addresses major challenges notably performance requirements: < 75 µs latency (including fibre transmission which adds 5 µs/km), < 8 ns timing error, and several tens of Gbps throughput for critical signals, values which are outside of the current work. It extends deterministic networking across multiple stakeholders and domains, and the associated strict reliability requirements associated with such networking in dynamic environments should be covered. Key properties of the wireless communication link are considered as an integral component for the overall deterministic network design.

Proposals may address one or more of the topics above.

[1]meaning: both conceptually and in operations, i.e., with isolation and guaranteed quality levels