Agile and Cognitive Cloud-edge Continuum management

The AC3 project will rely on AI/ML techniques to build application profiles and contexts that will answer questions such as: What type of traffic? When and where are the traffic and data generated? The profile will then be used by the CECCM to predict the suitable placement of the microservices on the CECC infrastructure or trigger the appropriate LCM actions (i.e., migration or duplication), anticipating resource outage due to, for example, a potential data avalanche or high CPU usage. The profile specification will leverage web semantic and ontology to ease the adaptability of LCM actions to take the right decisions on the application and CECC infrastructure and adapt according to the changes in the users need.

The AC3 project will employ an AI/ML algorithm to predict CECC resources (i.e., Cloud/ Edge Computing resources, networking, and Data) usage as well as far edge availability, which when combined with application profiles, will help determine the optimal placement of the microservices that will run the application on the CECC infrastructure. The application profiles will be defined using a novel semantic-aware and ontology templating language (SOTL). The placement of the microservices will use trained models to ensure load balancing in order to optimize energy consumption while considering applications SLA. The AC3 project will fundamentally support energy consumption optimization thanks to the use of the microservice paradigm and the autonomous capability for taking decisions by employing ML and semantic techniques. Indeed, the AC3 project will use microservice migration as a key solution to consolidate the computing node of CECC infrastructure and reduce energy consumption. This is possible since migrating microservices have a very low impact on energy consumption than migrating monolithic applications. By leveraging the strength of AI and SOTL, the microservices running on top of CECC infrastructure can be optimally placed or migrated to available green nodes in an autonomous without affecting the SLA. By green nodes, we refer here to DC or hosts that use green energy instead of brown energy.

The AC3 project aims to guarantee agile LCM operations by dynamically migrating and duplicating microservices where necessary over the CECC infrastructure according to events. Therefore, a novel networking solution will be envisioned to update routes and networking resources to guarantee SLA dynamically. The AC3 project will envision solutions such as Computing First Networking (CFN), which considers cloud/edge resources as a criterion for routing decisions. The AC3 project will leverage CFN by using Software Defined Networking (SDN) on top of the CFN architecture to add dynamicity and flexibility when managing the traffic flow destined to the microservices that have been migrated in the CECC infrastructure, while ensuring SLA of applications.

Last but not least, AC3 will natively include data management procedures, by integrating these procedures in a Platform as a Service (PaaS)-like component to help the application developers to build applications that use stored or live data (stream retrieved from sensors, or logging modules). The AC3 data management PaaS will cover all data management procedures, such as data indexing, searching and retrieval, parsing, storing, transferring, managing, monitoring, streaming, etc. To ensure seamless access and management of microservices and data sources while reducing the interaction between different actors (e.g., users, applications, and data sources), SOTL will be leveraged for generating inquiries and policies that have the ability to adapt according to different contexts.

It is worth mentioning that AC3 will make considerable efforts to ensure the explainability of the used ML models, which will increase trust in AI/ML outputs and improve the decision-making process to manage the CECC infrastructure.