• Christine Banek (LSST)

SciServer is a platform user oriented that implements resource sharing, agnostic storage to host and share datasets, query and analysis tools. Is can be used as collaborative workspace. It can be installed with K8S and HELM. It can be used to explore CPU and GPU computing or to access DASK or SQL Server 2019 Big Data Studio.

Q: how can I make my SP interoperate? in practice how can a user execute a contaner based on my image on another science platform?

Q: sustainbility problem, I have a finite resourcepool, can I make my SP able to deploy on commercial clouds so that users pay for it? Is it a functional sustainability approach?

STScl open source oriented science platform that profit of github and AWS clouds. It uses K8S HELM and Docker and provide Jupyterh Labs to users.

Q: Security How do we secure the environment from external and internal sources? How do we control exclusive access data in the environment?

Q: Collaboration How do we provide an environment that enables collaboration while being secure?

Q: Observability What are the important metrics to monitor? How do we have real time insight into the system?

Q: Cost How do we control costs for different users and different use cases? How do we provide access for different users?

CDS Science platform concepts A science platform could provide an access to Simbad and VizieR through APIs, visualisation through Aladin Lite (ipyaladin) and python tools for HiPS and MOC, computational facilities to X-Match catalogues.

Comment: It must be well framed and sized, complementary to the services and developed with scientists for scientists => scientific sustainability

Q: storage space is necessary and must be allocated

Q: Can we accept external Containers?

LSST science platform allows accessing to data and processing for LSST. There are three interactive ways of exploring the data portal (discovery, structured work flow between datasets, querries), notebook environment (ipython, contemporary hub technologies) and public API/VO and REST.

VO-CLOUD is an example of science paltform that implements batch processing based on UWS VO standard and it is focusing on interactive Machine learning: "active learning".

The European funded project ESCAPE is developing a science platform on top of European Open Science Cloud. It will include VO data access standards, AARC compliant authentication and authorization, containers from the ESCAPE software market place and EOSC market-place.

Q: what can IVOA provides to simplify the design and implementation?

The Virtual Observatory point of view (not IVOA) identify some important items to build real interoperability and in particular what is missing and what is already available:

• data discovery and access, open VO standards
• interoperable AAI solutions (IVOA GWS working on it ... )
• metadata to characterize network proximity
• IVOA metadata EXIST for describing data, NEED to have them for code
• there probably already are some standards for this
• we should find/evaluate and adopt them rather than invent new ones

The use of OAuth tokens for authentication has been discussed by IA2 Sonia Zorba that introduce the problems related to the use of tokens outside the browser and web enviroment. We must support also CLI and terminals so this is crucial for Astronomers.

Q: should we build complex clients or use separate AA for CLI?

Q: what about using certificates? PKI is still very supported and secure.

Conclusions

We identify three different science platforms types:

• Applications oriented: platform that provides access to tools and visualization
• Project oriented: access to data and sw and pipelins and contaners for a specific Project.
• user oriented: the user can also provide his own software and data.

• Interoperability among science platforms aims at enabling direct re-use of data/service/code/workflow resources among standalone platforms
• "ecosystem is better than monoculture" so we should build iteroperability from the ecosystem idea
• Interoperability means:
  • interoperable data resources
  • interoperable code resources
  • moving them around in a realistic and reasonable way
• The use of FAIR principles

Actions:

• Common Resource access control?
• Run my platform on your k8s cluster?

We identify research area for the future: