July 2017 GDB notes

Agenda

https://indico.cern.ch/event/578988/

Introduction (Ian Collier)

presentation

Upcoming GDBs

no GDB in August. September pre-GDB on Storage. March 2018 to be co-located with ISGC.

Pre-GDB on HPC Resources

Goal create a forum for LHC experiments and HPC systems teams. Collect best practices. Pathfinder to integrate resources. Will be held 2nd February 2018.

Discussion: IPv6 for April 2018

we would like WN only IPv6 working seamlessly. Plans will be made in coming weeks to start campaign to move sites. Some sites will be ticketed. There is a HEPiX IPv6 meeting this week to discuss the process. There is also a twiki page that sites will be asked to update with their status.

HNSCiCloud update (Bob Jones)

presentation

• Reminder – it is a joint pre-commercial procurement. Budget just over 5M euros.
• Looked at needs. Common tender. Bids evaluated. 4 designs evaluated and 3 selected for prorotyping. 3 in this stage. Next pilot.
• Research community is not just HEP. Focus is IaaS.
• Challenges: Compute & storage; network connectivity & federated identity management and service payment models.
• There are 3 consortia implementing the prototypes.
• There test phases in an overall test plan. Cores and storage ramps up during each phase. First part was establishing (e.g. network links to Geant).
• See https://cernbox.cern.ch/index.php/s/pjy4lg3cY46vbGt
• Next steps: 2nd progress review 4-5 September; 3rd progress review 9-10 October.
• Pilot phase 2018 will take 2 prototypes into pilot phase.
• Currently 10 organisations. Will establish early adopter group.

Q: Is there much sharing between the 3 consortia, for example to cover common issues that are arising A: There are

Q: When will this have to be realized? Bob: Will look at this in 2018. Formalise buyers group to become the formal mechanism for purchasing later. See the value of linking with the Geant model. All of this is voluntarily.

Q:This is a European project. If we are to use it in WLCG how difficult will it be for others to join?

A: From a technical perspective not difficult. Standard protocols often used. Questions about procurement model will need to be resolved – legislation and procurement rules will come into play from other regions. Building a base procurement framework and there could then be different procurement groups to bring in different regions.

Authorisation Working Group proposal (Liviu Valsan)

presentation

• Motivation: Evolving identity landscape. Central user blocking. (EU) Data protection.
• Objective: Understand and meet the requirements of an authorization service for the WLCG VOs.
• Stakeholders: WLCG security; VO management & service developers…
• Plan: Understand usage -> identify limitations -> Identify solution.
• To get involved: e-group project-lcg-authz. Pre-GDB in November.

Comment: Important to clarify the language. We are talking about suspending credentials (not just banning an individual).

Q: Is the timeline sufficient if first meeting is in mid-November? A: There will be virtual meetings before then.

Q: Difficulty banning the VO credentials is impractical. Am I mixing things? This is not SSO vs x.509. It has to do with the VO auth in VO land or site land. VOs said trust us. The agreement was that a VO could then be banned. A: Not always helpful for security – how would you manage the unbanning stage? Better to have access control and suspension control with sites so they can use it when they need it.

Q: Agree VO banning is a last resort. Not the same as saying it is difficult. A: Extremely difficult to undo. C: All this time when glexec partially accepted we have been in this situation. More a political issue.

Q: Can the presentation be updated with these more accurate points? Afterall it is the record.

Q2: Want to understand the scope of the WG better. How does it relate with the efforts in recent European projects – e.g. in IndigoData Cloud provide a system to do many of the things described. November is rather close. Share the concern about the terminology. What should be clear in objectives – what has changed that requires more thinking in this area given that we have already got tools for suspending (e.g. in VOMS).

A: Yes have looked at works in recent European projects. Mostly for EGI where there is a large number of VOs. The requirements may be different to 4 large VOs. C: Do not conflate EGI and IndigoDataCloud. You mentioned review and solve but not assessing other options. A: Yes we will look at existing solutions. Comment: Surveying what others are doing then is perhaps an additional explicit step. Comment 2: WLCG is a collaboration of infrastructures that have their own plans for evolving. We need to make sure experiments can continue to access these seamlessly. The goal is not when we expect to have this all sorted, it is just a good point at which to get together for a face-to-face discussion, not to give recommendations. A: Hope other stakeholders will participate in November.

Comment-EGI: The plan is to use approaches to access resources that are not only accessed by WLCG VOs, so the stakeholders must also include the sites. A: The latest slides already mention “all welcome” it was not meant to be an exhaustive list.

Comment: Not just Indigo working in this area but also ARC and ARC2. Hannah is involved in those.

Database Workshop Report (Eva Dafonte Perez)

presentation

• Workshop took place 29th-30th May. Slides in indigo.
• Aims: Discuss future requirements; Identify common needs; Evaluate trends; Understand service evolution.
• 1st workshop in this series was held 6 years ago.
• Tracks: Requirements run 3&4. Implementations & technologies. Going beyond relational.
• Next workshop perhaps 2019 (LS2).

Q: Oracle is still there. What about costs? Was that factor discussed? A: No. The cost of moving everything would be a lot. We are offering many solutions so users can start moving. Also currently renegotiating contract.

Q: What about participants from other sites? A few T1s have to run the instances. Were there any comments? A: Some sites covered, but some will not even deploy it.

Comment: Nobody is proposing to introduce new Oracle services. For CERN the cost of physics databases is not an issue. There is little left than that that needs to be there – more now on mySQL.

Container Technologies pre-GDB summary (Gavin McCance)

presentation

• Ian C: the future use of Argus also is a matter for the new Authorization WG

Manchester Workshop report (several speakers)

presentations

• Q: Won't SKA already do a lot of processing locally at their observatory sites?
• A: Indeed, but they still send huge amounts of derived data to their regional centers,
  which is what matters for common projects, frameworks and services

• Q: Was the end of Globus support (gftp) discussed?
• A:
  • A little. But more was discussed at the HSF workshop the following week.
  • OSG and CERN have agreed to take on support.
  • Contacts for people involved shared but nothing concrete done for work allocations at this stage.
  • The basic message is we do not need to worry about it for now. We may get better support.
  • There is also an ongoing discussion within PRACE.
    • The first person to speak to within WLCG is Oliver. MJ to send an email to Ian.

• Q: How precise are the network requests based on experiment input?
• A: Not very precise, and only for the medium term.

• Michel: StackExchange requires many people to have the right reputations
• Markus: at least the model can be used
• Jeff (chat): try something like http://distill.pub/journal/ ?

WLCG recommendations for batch system and CEs (Erik Mattias Wadenstein)

presentation

• Mattias and Maarten will follow up
• to be discussed further in the Sep GDB

Next WLCG Workshop (Ian Collier)

presentation

• Apr 25 holiday not a problem for Bari, university facilities would be available
• candidate sites are invited to prepare short presentations for the Sep GDB
• the HSF still has to confirm to a combined workshop