Project deliverable Open Access

D5.9 Service deployment and cloud computing on distributed Nordic cloud resources

Neukirchen, Helmut; Erlingsson, Ernir; Janos Szentannai, Lorand; Pica, Claudio; Thrane, Dan Sebastian; Obst, Matthias; Deneke, Tewodros; Azab, Abdulrahman

The goal of the T5.2.3 subtask is to investigate and showcase innovative platforms for making scientific tools able to discover and consume cloud computing resources, such as IaaS and Kubernetes solutions running either on-premises and/or on public clouds.

The subtask investigated solutions provided by Sigma2’s NIRD Toolkit1, STACKn2, and FEDn3 from Scaleout4, cPouta5 from CSC, as well as UCloud6 from SDU. These provider solutions were selected based on the user experience and eligibility to host and analyze the associated research data. The goal is to provide a platform for researchers to deploy and run a variety of tools on user-selected data and cloud computing resources, giving researchers the possibility to focus mainly on research and less on tools and irrelevant tasks. The solutions we have investigated as part of the subtask are to enable AI/ML workflows for natural language analysis, climate modeling and biodiversity use cases and to provide cloud infrastructure resources as well as to enable sharing of workflows across borders.

Among tasks and activities, STACKn (formerly known as LeanAI) was brought into EOSC-Nordic. STACKn was deployed on Kubernetes resources at Sigma2 on the NIRD Service Platform (NIRD SP), and at Scaleout. This gave opportunities for demonstrating cross-border cloud computing capabilities and facilitating dialog for collaboration with the NLP community.

The NLP community has channeled its focus and efforts into consuming more traditional HPC resources. Since slightly different needs arose from the marine biodiversity community, FEDn was chosen as the technical platform solution.
The task investigated possibilities for enhancing the interoperability between cloud container-based computing and the traditional HPC to support compute-intensive AI/ML workflows from the EOSC-Nordic research community. A proof concept based on UCloud has been developed, deployed, and thoroughly tested on the NIRD Toolkit.

Furthermore, this subtask discussed and evaluated possibilities for setting up a Nordic cloud toolbox, where other service providers would have one single entry point for accessing all the tools and cloud infrastructure enablers in one place.
Finally, an effort has been invested into prototyping accounting for better cloud resource distribution, optimal utilization, and improved resource planning and prediction.

After a close dialogue with communities, three main use cases have been investigated and worked on to develop solutions that can support research, facilitate collaboration and scale out of silos offering scalable cloud computing solutions.

This is the draft version of the deliverable not yet approved by the European Commission.
Files (10.7 MB)
Name Size
EOSC-Nordic D5.9.pdf
md5:3badc729b5ca36d76dd9e63b39ce5927
10.7 MB Download
18
16
views
downloads
All versions This version
Views 1818
Downloads 1616
Data volume 171.2 MB171.2 MB
Unique views 1818
Unique downloads 1616

Share

Cite as