— WP objectives
Leader: Walid Gaaloul
- Design and implement an orchestration architecture for resources, services, and applications across the cloud edge IoT continuum.
- Design and implement multi-layer composition of orchestration functional properties
- Develop adaptive and autonomic orchestration algorithms considering network partitioning
- Explore opportunities for cooperation and coordination between autonomic loops operating at different levels of the IoT-Edge-Cloud continuum.
— Missions
Orchestration is the process of dynamically and automatically managing Cloud/Edge/IoT resources, applications and services to meet end-user needs. This work package will be linked to the monitoring and analysis solution developed in the PEPR Cloud SPIREC project, which aims to predict resource demand, server failures, etc. These elements are needed in our project to build an autonomous orchestration solutions to proactively allocate the necessary resources to ensure optimal availability of services in an orchestration. The SLICES-FR platform, to which the SILECS project contributes, will be used for the experiments. The models proposed in WP1 for configurable and decomposable orchestration will also have to be considered by the orchestration engines in this WP. We will also use the provisioning and configuration languages and reconfiguration techniques resulting from WP2 to build efficient and secure orchestrators. Our orchestration will also build on the results of WP4 to optimize orchestration performance (e.g., execution cost and energy consumption) by providing a framework for integrating their optimization techniques.
— Tasks description
T3.1 Multi-level management of functional properties that address the network topology of the cloud-edge-IoT continuum
The orchestration of services and resources in the IoT-Edge-Cloud continuum is characterized by two challenges related to the orchestration topology in the network: (i) the complexity of the granularity levels of IoT-Edge-Cloud services and resources, and (ii) the network partitioning that affects the smooth operation of applications in disconnected mode. To address these challenges, we are developing an orchestration architecture that considers intent in the form of capabilities to describe the functional dependencies of services in the topology (see Task 1.3 in WP1). In addition, the move to the IoT-Edge-Cloud continuum raises the issue of network partitioning becoming the norm, and therefore the need to explore new approaches to ensure the smooth operation of applications in disconnected mode. Implementing this architecture will require the development of new software components, which will be extended to existing Cloud/IoT service orchestrators such as Kubernetes.
T3.2 Autonomic orchestration in the Cloud-Edge-IoT continuum
Autonomic management and orchestration of distributed systems involves control loops to handle a variety of objectives (e.g., self-optimization, configuration, protection, etc.). In this context, we highlight two challenges:
- The first is the need to coordinate multiple, even parallel, loops to address the coexistence of multiple problems in complex real-world architectures. A complementary challenge is the introduction of concurrency in MAPE-K (Monitoring/Analysis/Planning/Execution-Knowledge) autonomous loops. This goal is complementary to Task 2.2, one of whose goals is to enable concurrency of execution phases. There are also links with WP4 to define appropriate placement algorithms.
- A second challenge concerns the application of machine learning techniques of a generally centralized nature to the construction of the autonomous loop in the distributed environment of the IoT-Edge-Cloud continuum. By interfacing with the SPIREC project to detect and predict anomalies, we aim to proactively act on the autonomic loop to provide the necessary resources and patches to maintain optimal orchestration availability. Experimental validations will be performed on the SLICES-FR platform to ensure reproducibility with respect to T2.3 and T3.4.
T3.3 Managing non-functional orchestration properties in the Cloud-Edge-IoT continuum
In Task T3.3, which is linked to Task T3.2, we focus on some important non-functional properties of orchestration in the IoT-Edge-Cloud continuum.
- The first is the vulnerability of the IoT-Edge-Cloud continuum to privacy attacks, and in particular to Byzantine participants, despite the decentralization of data in federated learning. We therefore plan to improve this learning in collaboration with the TRUSTINCloudS project.
- The second property is energy footprint management, where most work on the energy footprint of the cloud considers an application as a “black box” unaware of the underlying infrastructure. However, closer cooperation between the hosted application and its infrastructure can lead to interesting results. For example, the infrastructure could dynamically inform the application of its carbon footprint and the energy constraints of the infrastructure, allowing the application to adapt to manage its QoS in a controlled manner.
T3.4 Deploying basic infrastructure and services for Cloud-Edge-IoT
The objective of this task is to develop a service and infrastructure provisioning framework that can manage the configuration, provisioning, and management of cloud and hardware resources, applications, and services to ensure business continuity. The links between service provisioning, as proposed in Task 3.2, and infrastructure management are closely related to physical platforms and intended uses. Greater versatility in infrastructure management means that more comprehensive services can be envisaged and that non-functional characteristics can be more easily taken into account. The approach is intended to be complementary or even alternative to T3.3 by enabling and exploiting richer low-level or first-level deployments. In addition, the orchestration framework must include a large-scale asynchronous communication service that can operate in disconnected mode and support other TARANIS core services or PEPR applications, or higher-level application services on the IoT-Edge-Cloud continuum. Experimental deployments will be performed on the SLICES-FR plateform.
Other Work packages
