Motivation
Currently, three LHC experiments have deployed or are ready to deploy a federated model for experimental data access (i.e., ALICE, ATLAS, and CMS). While each experiment has taken a similar approach (i.e. xrootd), the implementations are sufficiently different to warrant a collaborative discussion of the specific needs that spurred a particular direction.
Purpose
Through a collaborative dialogue, this meeting hopes to expedite adoption and support of federated data access models in LHC by leveraging each other’s experience.
Points of discussion
- Federated Data access models:
- Mechanisms:
- Copy on demand.
- Real time access.
- Partial caching.
- Pros and cons to each mechanism:
- Requirements to make each model work well.
- Firewalls.
- Federated Name Space:
- Providing a uniform global logical name space:
- Challenges and solutions.
- Federated Monitoring:
- Metrics:
- Necessary information.
- Desired information.
- Level of detail.
- Visualization:
- Incorporating local information:
- Mechanisms.
- Useful metrics.
- Federated Security:
- Requirements for different access modes:
- Client/Server vs Server/Server access.
- Emerging mechanisms:
- DOE’s InCommon Federated Trust Model, Shibboleth, etc.
- Federated Support Structure.
Meeting Format
We envision a 20-30 minute technical and practical presentation in each topic area from each group (experiment or computing group) developing or deploying a federated data access model and related services. The presentation should include a description of what worked and did not work and why; as well as challenges yet to be addressed.
