ESCAPE WP2 Fortnightly 16th May 2019
Participants: Simone, Gonzalo, Rosie, Aris, Mario, Xavier, Manuel, Tommaso, Ghita, Paul, Andrea, Simone, Frederic, Ron
- WP2/WP5 joint meeting Amsterdam (1-3 July 2019) agenda is ready https://indico.in2p3.fr/event/19214/ (registrations not yet opened but timetable confirmed)
- At the next WP2 fortnight meeting will dedicate time to discuss the topics for the face to face meeting in Amsterdam.
Datalakes ideas from FAIR (Kilian Schwarz)
- The basic ideas is:
- Global access point and redirect user by geo-IP to local endpoints
- Squid proxy entry to data lake
- Dynafed to aggregate different storages in one namespace and communicate via http to Dynafed
- Sci-token are foreseen
- Groningen site under discussion experience with iRODS for metadata, process of roll out the datalake idea
- Auth server for sci-token and replication manager xrootd base framework
- Q(Xavi): How the client based geo-ip mapping and Dynafed work together?
- A(Kilian): Good point need to investigate how this two can work together… usage of squids might be useful.
- Q(Paul): How this works?
- Q(Mario): Done this in ATLAS, will circulate papers.
- Q(Paul): How the user will authenticate in the Sci-Token server. Need to be an authentication: Open-ID connect or x509
- A(Kilian): No idea yet. SAML is a possibility.
- Q(Andrea): What is the auth/authz model? by groups?
- C(Simone): WP2 a work package on AAI, need to have a coherent approach. Roles or capability based? etc. How flexible you are in using Sci-Tokens? and tools for data replication: FTS or equivalent third-party tools?
- A(Kilian): Sci-Tokens is an idea as it seems to fulfil requirements but we are flexible. For the replication want to start with simple approach.
- C(Mario): Implementing Dynafed on top means to limiting access to Webdav and S3.
- A(Kilian): looking at xrootd forward proxy and allow xroot protocol
- C(Paul): Usage of http in TPC transfers can be done in xroot (ie. FTS)
- Q(Simone): Which sites are involved?
- A(Kilian): GSI, Groningen, SUT and KIT
- Q(Simone): What about caching and latency hiding? which data access patterns you expect?
- A(Kilian): We expect remote data access. System redirect the user to the closest data. Our system was the inspiration for XCache.
- Q(Manuel): Different variation for http data access: suppose my data is somewhere we can only talk http and I want to access it via xrootd, then the cache need to talk http but serve it via xrootd.
- A(Mario): Alistair suggested have permanent solution on cloud, go http and then Dynafed redirects and give another protocol. The other idea is do this client based: client requests a protocol translation to a service like FTS that will stream it with the correct protocols.
- Q(Manuel): Instead of streaming we could think about pin things to the cache and when doing the pinning this is only http.
- C(Paul): Theoretically possible and IIRC hinted in some presentations I’ve seen.
- C(Mario): Remember meta-links to list all available protocols
Update on the hiring process:
- CERN: In the middle of hiring process, we will have a person on June 1st and one on July 1st if all goes well.
- CNAF: starting hiring senior fellow at CNAF for AAI
- GSI: advertised the two slots they should be in the web soon, process is launched.
- CCIN2P3): Still considering if we can work with someone that is already on site or open a temporary position.
- SKA: Agreed two hires, signed paperwork WP2 and WP5. The post for networking is not yet advertised (12person/month) as we want to make the networking post as a full post, so there is a bit of delay about that.
Next meeting: TBD as 30th of May is holiday at CERN, will distribute a doodle poll.
There are minutes attached to this event.