Periodic Reporting for period 1 - ACES (Autopoietic Cognitive Edge-cloud Services)
Reporting period: 2023-01-01 to 2024-06-30
ACES will solve challenges by infused autopoiesis and cognition on different levels of cloud management to empower with AI different functionalities:
i) autopoiesis cognitive cloud-edge framework;
ii) awareness tools, AI/ML agents for workload placement, service and resource management, data and policy management, telemetry and monitoring;
iii) agents safeguarding stability in situations of extreme load and complexity;
iv) swarm technology-based methodology and implementation for orchestration of resources in the edge;
v) edge-wide workload placement and optimization service;
vi) an app store for classification, storage, sharing and rating of AI models used in ACES.
To foster the uptake of ACES outcomes beyond its lifespan, different activities are foreseen to drive adoption to a wider network of stakeholders in key sectors.
i) autopoiesis cognitive cloud-edge framework;
ii) awareness tools, AI/ML agents for workload placement, service and resource management, data and policy management, telemetry and monitoring;
iii) agents safeguarding stability in situations of extreme load and complexity;
iv) swarm technology-based methodology and implementation for orchestration of resources in the edge;
v) edge-wide workload placement and optimization service;
vi) an app store for classification, storage, sharing and rating of AI models used in ACES.
To foster the uptake of ACES outcomes beyond its lifespan, different activities are foreseen to drive adoption to a wider network of stakeholders in key sectors.
WP2: Crafted the architectural blueprint for the ACES project, meticulously considering the diverse requirements of its use cases.
WP3: Researching and defining knowledge capture techniques to enhance agents' reasoning for actions in WP4 as well as the technological foundation to support efficient knowledge acquisition, generation and storage.
WP4: Researching interactive actions using swarm intelligence/gradient tracking/bayesian learning for workload placement, and its optimization within an autopoietic context leading to a first conceptualization of possible candidate algorithms to the ACES architecture
WP5: Creation of a plan to develop and test CUs with the relevant platform services; initial methodology for validating and demonstrating the pilots using the ACES platform; initial data was retrieved from SCADA systems and network topology; procurement of EMDC servers is progressing; the use cases have been developed by 80% and shared with the relevant partners on the basis of non-disclosure agreements.
WP3: Researching and defining knowledge capture techniques to enhance agents' reasoning for actions in WP4 as well as the technological foundation to support efficient knowledge acquisition, generation and storage.
WP4: Researching interactive actions using swarm intelligence/gradient tracking/bayesian learning for workload placement, and its optimization within an autopoietic context leading to a first conceptualization of possible candidate algorithms to the ACES architecture
WP5: Creation of a plan to develop and test CUs with the relevant platform services; initial methodology for validating and demonstrating the pilots using the ACES platform; initial data was retrieved from SCADA systems and network topology; procurement of EMDC servers is progressing; the use cases have been developed by 80% and shared with the relevant partners on the basis of non-disclosure agreements.
Knowledge graph SOTA: often focuses on using graph theory to optimise resource allocation, load balancing, and data flow in such systems cloud computing systems, yet are collecting this information in a centralised location from which in a hub and spoke set-up decisions are being made.
BSOTA: ACES goes beyond existing research by introducing a new graph model that is installed locally on each of the locations in the edge-cloud infrastructure where it holds a local and partially 'global' view.
BSOTA: This work might extend the current state of research by defining dynamic and intelligent metrics acquisition flows that are adaptive, context-aware, and capable of triggering optimisation's autonomously.
BSOTA: Bayesian methods are used for probabilistic reasoning and adaptive learning and allow agents to to make informed decisions under uncertainty and continuously update their knowledge based on new information.
Load balancing SOTA: Servers (CPU cores, DRAM and a fixed attached GPU and storage) are currently the smallest unit of compute in the resource allocation and load balancing.
BSOTA: The ACES framework collects a more fine-grained understanding of resources and workloads.
BSOTA: swarm intelligence i used to match supply of resources dynamically created from pools of resources, with the demand for resources. Swarm intelligence dynamically schedules and load balances and a knowledge graph and AI is being used to understand the swarm parameter settings and the effect they have on the platform KPI e.g. energy efficiency, utilisation levels, meeting SLA's, SLO's.etc.
BSOTA: In order to create AI governance for the infrastructure operator an app store for classification, storage, sharing and raying of AI models is developed.
Key needs for commercialisation, internationalisation. In order to guarantee further uptake:
- opens source the framework for decentralised autonomous edge cloud orchestration.
- integrate with the CloudEdgeIoT standardisation effort that is currently undertaken by the EC and the tech community in Europe.
- create a large demonstrator, distributed edge cloud infrastructure where the framework is being deployed and further developed by the community. Telco edge cloud would be a good demonstrator and focussed on a specific industry vertical; smart healthcare, intelligent energy grid, etc.
- create a standard topology for edge cloud hardware for edge datacenters, based on Industry standard com-express and com-hpc and including cxl. European OEM Industry is well represented in manufacturing under this form factor standards. CXL will rapidly be implemented and dominating the cloud datacenter industry and unless this uptake stretches also into the topology of edge-cloud infrastructures the dominance and lock-in from cloud providers will become even more challenging. By developing our European standards for the de-cloud infrastructure, in line with OCP initiative, we create a head start for European Industry and at the same time speed up the development of energy efficient edge data centers that is now being held back by large Industry.
Telco edge cloud is currently the largest joint effort from an industry vertical compared to smart health, farming, energy grid, etc. All these other sectors lack the widescale and joint efforts and investors to build infrastructure. Currently the Telco industry is making strategic deals with the large cloud providers, often non European. When we can bring European Telco on a European defined and standardised infrastructure that is rolled out together with 5G and 6G, the first milestone of a European infrastructure will be set. If European Telco's will adopts cloud service providers technology standards and hardware, the edge cloud as such will have a standard that is not European, nor designed and adopting European values, of sovereignty, energy efficiency, etc. Commercialisation of European Edge cloud on European technology will then become very unlikely given the strategic importance of telco.
BSOTA: ACES goes beyond existing research by introducing a new graph model that is installed locally on each of the locations in the edge-cloud infrastructure where it holds a local and partially 'global' view.
BSOTA: This work might extend the current state of research by defining dynamic and intelligent metrics acquisition flows that are adaptive, context-aware, and capable of triggering optimisation's autonomously.
BSOTA: Bayesian methods are used for probabilistic reasoning and adaptive learning and allow agents to to make informed decisions under uncertainty and continuously update their knowledge based on new information.
Load balancing SOTA: Servers (CPU cores, DRAM and a fixed attached GPU and storage) are currently the smallest unit of compute in the resource allocation and load balancing.
BSOTA: The ACES framework collects a more fine-grained understanding of resources and workloads.
BSOTA: swarm intelligence i used to match supply of resources dynamically created from pools of resources, with the demand for resources. Swarm intelligence dynamically schedules and load balances and a knowledge graph and AI is being used to understand the swarm parameter settings and the effect they have on the platform KPI e.g. energy efficiency, utilisation levels, meeting SLA's, SLO's.etc.
BSOTA: In order to create AI governance for the infrastructure operator an app store for classification, storage, sharing and raying of AI models is developed.
Key needs for commercialisation, internationalisation. In order to guarantee further uptake:
- opens source the framework for decentralised autonomous edge cloud orchestration.
- integrate with the CloudEdgeIoT standardisation effort that is currently undertaken by the EC and the tech community in Europe.
- create a large demonstrator, distributed edge cloud infrastructure where the framework is being deployed and further developed by the community. Telco edge cloud would be a good demonstrator and focussed on a specific industry vertical; smart healthcare, intelligent energy grid, etc.
- create a standard topology for edge cloud hardware for edge datacenters, based on Industry standard com-express and com-hpc and including cxl. European OEM Industry is well represented in manufacturing under this form factor standards. CXL will rapidly be implemented and dominating the cloud datacenter industry and unless this uptake stretches also into the topology of edge-cloud infrastructures the dominance and lock-in from cloud providers will become even more challenging. By developing our European standards for the de-cloud infrastructure, in line with OCP initiative, we create a head start for European Industry and at the same time speed up the development of energy efficient edge data centers that is now being held back by large Industry.
Telco edge cloud is currently the largest joint effort from an industry vertical compared to smart health, farming, energy grid, etc. All these other sectors lack the widescale and joint efforts and investors to build infrastructure. Currently the Telco industry is making strategic deals with the large cloud providers, often non European. When we can bring European Telco on a European defined and standardised infrastructure that is rolled out together with 5G and 6G, the first milestone of a European infrastructure will be set. If European Telco's will adopts cloud service providers technology standards and hardware, the edge cloud as such will have a standard that is not European, nor designed and adopting European values, of sovereignty, energy efficiency, etc. Commercialisation of European Edge cloud on European technology will then become very unlikely given the strategic importance of telco.