Periodic Reporting for period 1 - Ether (Development ecosystem intended for engineers and researchers looking to invent, build, verify, evaluate, and release application-tailored storage systems faster and more reliably)
Reporting period: 2020-05-01 to 2022-04-30
What is the problem/issue being addressed?
The Action “Ether: Development ecosystem intended for engineers and researchers looking to invent, build, verify, evaluate, and release application-tailored storage systems faster and more reliably” aims at improving the quality of IoT applications that depend on flexibility, scalability, functionality across large geographic areas, and data exchange across an entire system of systems.
Why is it important for society?
Autonomous vehicles, Smart houses, and Intelligent Healthcare are profound IoT cases grounded on the continuum of resources between centralized Cloud data centers and IoT nodes deployed in the field. In contrast to the well-powered and well-plugged nodes in data centers, the IoT nodes deployed in the field are frequently disrupted by power outages, poor connectivity, mobility, and several operating conditions which are constantly changing. Therefore, given modern society's exposure to such IoT applications, understanding their ability to cope with unforeseen conditions is critical to improving usability and user safety.
What are the overall objectives?
The objectives of Ether have been to:
1. Decompose existing systems and architectures and build an Atlas of primitives that collectively constitute a data management system for IoT.
2. Develop a fully automated Cloud-native platform for deploying, testing, and benchmarking IoT applications.
3. Build a collection of realistic scenarios that involve heterogeneous hardware resources, arbitrary network distances, and a broad range of failures (e.g. network delays, crashing nodes, disk errors).
A parallel goal of this MSCA Individual Fellowship was to foster the development of the individual researcher.
The Action “Ether: Development ecosystem intended for engineers and researchers looking to invent, build, verify, evaluate, and release application-tailored storage systems faster and more reliably” aims at improving the quality of IoT applications that depend on flexibility, scalability, functionality across large geographic areas, and data exchange across an entire system of systems.
Why is it important for society?
Autonomous vehicles, Smart houses, and Intelligent Healthcare are profound IoT cases grounded on the continuum of resources between centralized Cloud data centers and IoT nodes deployed in the field. In contrast to the well-powered and well-plugged nodes in data centers, the IoT nodes deployed in the field are frequently disrupted by power outages, poor connectivity, mobility, and several operating conditions which are constantly changing. Therefore, given modern society's exposure to such IoT applications, understanding their ability to cope with unforeseen conditions is critical to improving usability and user safety.
What are the overall objectives?
The objectives of Ether have been to:
1. Decompose existing systems and architectures and build an Atlas of primitives that collectively constitute a data management system for IoT.
2. Develop a fully automated Cloud-native platform for deploying, testing, and benchmarking IoT applications.
3. Build a collection of realistic scenarios that involve heterogeneous hardware resources, arbitrary network distances, and a broad range of failures (e.g. network delays, crashing nodes, disk errors).
A parallel goal of this MSCA Individual Fellowship was to foster the development of the individual researcher.
Work was conducted in three main directions: Research, Dissemination, and Exploitation.
In the research direction, the Fellow delivered two conference publications and one pre-print publication to date, with additional three forthcoming conference manuscripts. He was appointed External Reviewer to the Springer "Journal of Supercomputing". He was invited to serve as a Program Committee member at the "Testing Distributed Internet of Things Systems" (TDIS) workshop. He was invited by the MedPerf Task Force of the MLCommons Association to provide expert advice on systems testing for an open framework for benchmarking machine learning in the medical domain. To transfer knowledge, he conducted four presentations and co-organized two workshops for researchers, contributed to three internal projects, and provided mentoring for early career researchers. Finally, the Fellow spearheaded the preparation of a RIA proposal for the EU, driving the effort from building the consortium to the final submission.
For dissemination, the Fellow adopted Open Science practices; Publications drafts are made openly available on Zenodo, and the source code Frisbee platform is made available on GitHub (https://github.com/CARV-ICS-FORTH/frisbee) under Apache-2.0 License, accompanied by a DOI issued by Zenodo. Following the Fellow presented Frisbee in EuroCC webinars and high-profile open-source summits, both at the national (FOSSCOMM) and international level (FOSDEM). Finally, the Fellow joined the Marie Curie Alumni Association (MCAA) and performed as an MSCA ambassador at 2 European Researcher's Night events, demystifying the concepts of the Internet of Things and discussing with more than 300 high-school students the role of the researcher in today’s society.
Regarding exploitation, the Fellow's actions aimed at:
1. Engaging early adopters to validate Frisbee in real-life scenarios.
2. Finding resources for further research and development.
3. Commercializing the platform and the scenarios.
It is important to note, that through Ether and the project results, the Fellow has been gained visibility and has started to collaborate with corporate leaders in Telecommunications, Cloud storage, and Federated Learning. In addition, he incorporated the Ether concepts in two RIA proposals, aiming to extend Frisbee towards a tool for reproducible systems deployment and testing of complex, distributed IoT and datacenter applications. Finally, the Fellow has been in contact with investors to establish offerings, funding schemes, and viable routes for the commercialization of Ether results.
In the research direction, the Fellow delivered two conference publications and one pre-print publication to date, with additional three forthcoming conference manuscripts. He was appointed External Reviewer to the Springer "Journal of Supercomputing". He was invited to serve as a Program Committee member at the "Testing Distributed Internet of Things Systems" (TDIS) workshop. He was invited by the MedPerf Task Force of the MLCommons Association to provide expert advice on systems testing for an open framework for benchmarking machine learning in the medical domain. To transfer knowledge, he conducted four presentations and co-organized two workshops for researchers, contributed to three internal projects, and provided mentoring for early career researchers. Finally, the Fellow spearheaded the preparation of a RIA proposal for the EU, driving the effort from building the consortium to the final submission.
For dissemination, the Fellow adopted Open Science practices; Publications drafts are made openly available on Zenodo, and the source code Frisbee platform is made available on GitHub (https://github.com/CARV-ICS-FORTH/frisbee) under Apache-2.0 License, accompanied by a DOI issued by Zenodo. Following the Fellow presented Frisbee in EuroCC webinars and high-profile open-source summits, both at the national (FOSSCOMM) and international level (FOSDEM). Finally, the Fellow joined the Marie Curie Alumni Association (MCAA) and performed as an MSCA ambassador at 2 European Researcher's Night events, demystifying the concepts of the Internet of Things and discussing with more than 300 high-school students the role of the researcher in today’s society.
Regarding exploitation, the Fellow's actions aimed at:
1. Engaging early adopters to validate Frisbee in real-life scenarios.
2. Finding resources for further research and development.
3. Commercializing the platform and the scenarios.
It is important to note, that through Ether and the project results, the Fellow has been gained visibility and has started to collaborate with corporate leaders in Telecommunications, Cloud storage, and Federated Learning. In addition, he incorporated the Ether concepts in two RIA proposals, aiming to extend Frisbee towards a tool for reproducible systems deployment and testing of complex, distributed IoT and datacenter applications. Finally, the Fellow has been in contact with investors to establish offerings, funding schemes, and viable routes for the commercialization of Ether results.
This MSCA has pushed the frontiers of distributed and IoT systems specification, deployment, and testing in numerous ways.
1. The IoT emulator shed new light on the relationship between network links and node capabilities and provided guidelines on distributing services across the Cloud-IoT continuum. Meanwhile, Cloud-IoT paradigms can contribute to optimizing the energy and carbon efficiency of application deployments by: (i) Reducing the amount of data and the distance they must travel across the network; (ii) By exploiting idle edge devices to support application deployment; and (iii) accounting for the local environmental conditions, legislation frameworks, and operational costs on the deployment sites.
2. The Frisbee platform enhanced the emulator with a declarative methodology and event-driven concepts, passing the technological barrier of 'programming testing scenarios' and entering the age of 'describing testing scenarios' in almost natural language. Frisbee not only simplifies and makes test cases more explainable, but it also guarantees reproducibility. As a result, it can be equally used by industry for improving an application's performance/reliability/consumption, as well as by academia for conducting reproducible experiments.
3. Finally, this MSCA allowed the Fellow to explore additional disciplines and lead the authoring of a RIA proposal focused on streamlining the development, training, testing, and release of high-quality and privacy-preserving synthetic data generators for medical applications. Eventually, these generators have the potential to democratize AI research in Europe, as well as improve the accuracy and bias of existing clinical AI pipelines.
1. The IoT emulator shed new light on the relationship between network links and node capabilities and provided guidelines on distributing services across the Cloud-IoT continuum. Meanwhile, Cloud-IoT paradigms can contribute to optimizing the energy and carbon efficiency of application deployments by: (i) Reducing the amount of data and the distance they must travel across the network; (ii) By exploiting idle edge devices to support application deployment; and (iii) accounting for the local environmental conditions, legislation frameworks, and operational costs on the deployment sites.
2. The Frisbee platform enhanced the emulator with a declarative methodology and event-driven concepts, passing the technological barrier of 'programming testing scenarios' and entering the age of 'describing testing scenarios' in almost natural language. Frisbee not only simplifies and makes test cases more explainable, but it also guarantees reproducibility. As a result, it can be equally used by industry for improving an application's performance/reliability/consumption, as well as by academia for conducting reproducible experiments.
3. Finally, this MSCA allowed the Fellow to explore additional disciplines and lead the authoring of a RIA proposal focused on streamlining the development, training, testing, and release of high-quality and privacy-preserving synthetic data generators for medical applications. Eventually, these generators have the potential to democratize AI research in Europe, as well as improve the accuracy and bias of existing clinical AI pipelines.