Periodic Reporting for period 1 - SovereignEdge.Cognit (A Cognitive Serverless Framework for the Cloud-Edge Continuum)
Reporting period: 2023-01-01 to 2024-06-30
The project has 5 main goals:
1) Definition of a serverless framework architecture with AI-enabled disaggregated management:
The Consortium has made a first approach to an open architecture for FaaS-oriented Cognitive Continuum. The team identified functional components and interactions, with some initial testing. Among those components, the details of the AI-enabled orchestrator were further developed and its overall concept is being consolidated. In parallel, the initial efforts have been carried out to provide a private Cloud-Edge platform with FaaS functionality.
2) Develop a new distributed paradigm based on FaaS that provides access to the Cognitive Cloud:
The team made the initial steps towards an serverless model for edge application development that integrates with AI-enabled orchestrators.
Also, this includes the development of the components named Device Client and Provisioning Engine. The former includes both the internal code and the APIs. Both will be integrated with architectural components to support AI-enabled placement and updates.
Finally, the Consortium developed the Serverless Runtime that exposes information about function execution to integrate with AI-enabled orchestration.
3) Build an abstraction layer for secure and self-adaptive execution of FaaS across the Continuum:
The team has worked on the key concept of AI-enabled orchestration. It is about the development of formal layered models for the Cloud-Edge Continuum, a mechanism to continuously monitor and optimize infrastructure across multiple cloud/edge providers and finally the development of a model library for the selection and "hot swapping" between AI models.
4) New techniques for power-grid aware automatic and intelligent adaptation of the Continuum:
The team developed models and reasoning for energy efficiency and lower carbon computing. This relies on the use of a placement model/function across the Continuum based on predictive analytics for energy and carbon emissions.
5) Integration of AI-enabled Serverless Cloud-Edge Platform for deployment of new "continuum-native" applications:
The team created the infrastructure to enable CI/CD into a full COGNIT stack of the different components of the envisioned architecture. Different repositories contain the source code developed. One of them (“cognit-ops-forge”) automatically deploys the COGNIT stack either on-premise or on the public cloud. These open source repositories are accessible on public URLs under Apache 2.0 license.
Additionally, the team has produced an initial set of demos documented on the project's social media channels.
1) Definition of a serverless framework architecture with AI-enabled disaggregated management:
The Consortium has made a first approach to an open architecture for FaaS-oriented Cognitive Continuum. The team identified functional components and interactions, with some initial testing. Among those components, the details of the AI-enabled orchestrator were further developed and its overall concept is being consolidated. In parallel, the initial efforts have been carried out to provide a private Cloud-Edge platform with FaaS functionality.
2) Develop a new distributed paradigm based on FaaS that provides access to the Cognitive Cloud:
The team made the initial steps towards an serverless model for edge application development that integrates with AI-enabled orchestrators.
Also, this includes the development of the components named Device Client and Provisioning Engine. The former includes both the internal code and the APIs. Both will be integrated with architectural components to support AI-enabled placement and updates.
Finally, the Consortium developed the Serverless Runtime that exposes information about function execution to integrate with AI-enabled orchestration.
3) Build an abstraction layer for secure and self-adaptive execution of FaaS across the Continuum:
The team has worked on the key concept of AI-enabled orchestration. It is about the development of formal layered models for the Cloud-Edge Continuum, a mechanism to continuously monitor and optimize infrastructure across multiple cloud/edge providers and finally the development of a model library for the selection and "hot swapping" between AI models.
4) New techniques for power-grid aware automatic and intelligent adaptation of the Continuum:
The team developed models and reasoning for energy efficiency and lower carbon computing. This relies on the use of a placement model/function across the Continuum based on predictive analytics for energy and carbon emissions.
5) Integration of AI-enabled Serverless Cloud-Edge Platform for deployment of new "continuum-native" applications:
The team created the infrastructure to enable CI/CD into a full COGNIT stack of the different components of the envisioned architecture. Different repositories contain the source code developed. One of them (“cognit-ops-forge”) automatically deploys the COGNIT stack either on-premise or on the public cloud. These open source repositories are accessible on public URLs under Apache 2.0 license.
Additionally, the team has produced an initial set of demos documented on the project's social media channels.
Work performed:
- Analysis of the state-of-the-art infrastructure technologies to be used in the architecture.
- Requirement definition considering the perspective of the end users, the technology providers and needs related to sovereignty,interoperability,sustainability and security.
- Architecture definition: building blocks, interfaces and communication flows.
- Definition of metrics for decision-making within COGNIT.
- Collecting information to be used in the future validation, in particular to train AI models and perform stress tests.
- Generation of detailed tutorials and demonstration videos.
- ML models for workload characterization, and interference-aware placement to optimize energy usage.
- Development of a Cloud Continuum emulator, allowing to test the performance of the AI placement recommendation system at massive-scale.
- Creating architectures for each use case integration with the COGNIT Framework.
- Setting up and maintaining a testbed for validating and demoing the COGNIT Framework.
- Establishing software repositories and necessary infrastructure for the development and release of COGNIT open source prototypes.
- Processes, protocols, and tools for automating building and deploying the software solution for COGNIT.
- Each use case developed their own software integrations of their applications with the COGNIT Framework.
- Demonstration and validation of the integration of the COGNIT Framework in the use cases, and of the key functionality and performance aspects of the COGNIT framework.
Main achievements:
- Set of requirements to be met by the COGNIT Framework.
- Intermediate stable version of the COGNIT architecture.
- Methodology for architecture validation.
- Incremental versions of the different architecture components.
- Geo-distributed COGNIT testbed running the whole software stack.
- GitHub organization containing the project repositories.
- The OpsForge tool for automatically integrating and deploying the COGNIT stack.
- Integration and validation of the use cases with the COGNIT Framework with public repositories per use case
- Demonstration of the first version of the COGNIT Framework.
- Analysis of the state-of-the-art infrastructure technologies to be used in the architecture.
- Requirement definition considering the perspective of the end users, the technology providers and needs related to sovereignty,interoperability,sustainability and security.
- Architecture definition: building blocks, interfaces and communication flows.
- Definition of metrics for decision-making within COGNIT.
- Collecting information to be used in the future validation, in particular to train AI models and perform stress tests.
- Generation of detailed tutorials and demonstration videos.
- ML models for workload characterization, and interference-aware placement to optimize energy usage.
- Development of a Cloud Continuum emulator, allowing to test the performance of the AI placement recommendation system at massive-scale.
- Creating architectures for each use case integration with the COGNIT Framework.
- Setting up and maintaining a testbed for validating and demoing the COGNIT Framework.
- Establishing software repositories and necessary infrastructure for the development and release of COGNIT open source prototypes.
- Processes, protocols, and tools for automating building and deploying the software solution for COGNIT.
- Each use case developed their own software integrations of their applications with the COGNIT Framework.
- Demonstration and validation of the integration of the COGNIT Framework in the use cases, and of the key functionality and performance aspects of the COGNIT framework.
Main achievements:
- Set of requirements to be met by the COGNIT Framework.
- Intermediate stable version of the COGNIT architecture.
- Methodology for architecture validation.
- Incremental versions of the different architecture components.
- Geo-distributed COGNIT testbed running the whole software stack.
- GitHub organization containing the project repositories.
- The OpsForge tool for automatically integrating and deploying the COGNIT stack.
- Integration and validation of the use cases with the COGNIT Framework with public repositories per use case
- Demonstration of the first version of the COGNIT Framework.
The COGNIT Framework aims at becoming a reference solution addressing key challenges posed by a demanding and dynamic market. COGNIT will:
- Satisfy strong requirements like those about efficient data processing with low latency. This is relevant in applications like gaming, video streaming, autonomous vehicles, or smart cities and can be achieved through edge computing.
- Compete in the growing market of Serverless Cloud-Edge Frameworks, driven by technological advancements. It can be an attractive cost-effective solution with users paying only for the resources they use.
- Provide a Cognitive Cloud solution that will be part of the market of OS and orchestration components for Edge Computing, contributing to the efficient orchestration of distributed nodes.
- Propose a response in the context of the exponential increase in IoT devices demanding real-time data processing capabilities, which edge computing can provide..
- Process data as close to the source as possible, with benefit for security and privacy.
- Satisfy strong requirements like those about efficient data processing with low latency. This is relevant in applications like gaming, video streaming, autonomous vehicles, or smart cities and can be achieved through edge computing.
- Compete in the growing market of Serverless Cloud-Edge Frameworks, driven by technological advancements. It can be an attractive cost-effective solution with users paying only for the resources they use.
- Provide a Cognitive Cloud solution that will be part of the market of OS and orchestration components for Edge Computing, contributing to the efficient orchestration of distributed nodes.
- Propose a response in the context of the exponential increase in IoT devices demanding real-time data processing capabilities, which edge computing can provide..
- Process data as close to the source as possible, with benefit for security and privacy.