Date of Meeting: 4th March 2008

Attendees

DEISA: Hermann Lederer

OSG: Ruth Pordes (phone)

TeraGrid: Phil Andrews

NAREGI:

Invited: Dave Kelsey, David Wallom

Summary of AAAA Information

We agreed that we would make the information recorded in the spreadsheet public (but with better explanations of the entries based on Dave’s slides from the previous meeting) via an OGF managed wiki page with editing restrictions. Steven to set-up this wiki page.

OSG and EGEE already cooperate on security incidents. It is agreed that DEISA and TeraGrid will also cooperate and their security officers will determine how best to implement this security incident cooperation.

Dave has interacted with the JSPG to review AUPs and related policies and see whether they can be made to converge. He presented the www.jspg.org website, and specifically the Grid AUP: http://www.jspg.org/wiki/Grid_Acceptable_Use_Policy

Dave is waiting for feedback from TeraGrid (via Jim Basney) and still needs to consult with Jules Wolfrat (DEISA) to understand if there are any conflicts with their specific policies. This supplements local policies and to allow local users to use resources at other sites.

Review of resource allocation policies, procedures and portal information

The very complete Teragrid document is the only one that is easily available and locatable. There is significant divergence in terms of scope and David suggested that the scope should really be tailored to two specific audiences: Users and Resource providers for which different information and styles should be used.

In terms of criteria, there are significant difference between HPC (DEISA, TeraGrid) and HTC (EGEE, OSG). HPC has a more formal process involving a scientific and technical view of user requests while HTC relies on the resource providers input.

For portals, TeraGrid has the most developed approach to scientific gateways and even has a document called “How to write a successful science gateway bid”. David highlighted the following areas for potential alignment of the infrastructures:

• Projects and/or virtual organisations
  • Consider in the longer term allocation of resources dependant on measurable criteria? NGS in the UK is already doing this in places.
• This seems relevant for EGI but OSG is not considering following this path.
  • Current direct VO to resource centre relationship skews towards large VO
• In a sense there is a disjoint between HPC and HTC here: HTC accommodate more easily communities while HPC are tailored towards individuals
  • Users must be able to work out what resources they need beforehand
  • Communities cannot be individually reviewed (DEISA)
• Consider (please!)
  • common name
  • common structure where possible of the policy documents
  • common location for policy document
  • common focus, user community or resource provider etc

• Charging for utilisation
  • NGI resources will come from shared functionality resource operating under internal business models,
  • users must be able to show proof of ability to pay where necessary
• Co-allocation, new user communities requiring both HTC & HPC,
  • Common allocation units required
  • Must always be dependent on most stringent requirements
• Data only users,
  • David noted that, with the exception of TeraGrid, the resource allocation policies and procedures is very much focussed on CPU usage and that there is little attention

Next Meeting

In terms of participation, Bob will contact NAREGI to understand if they still wish to participate. Initial involvement was based on willingness to interoperate for the shared support of user communities. This basis should be used for any expansion of the group and also tied with some engagement at the technical level of OGF’s work. OGF27 will be held in Canada and that may be the opportunity to include the Canadian infrastructure.

Actions

• Steven will prepare a wiki page presenting the AAAA information
• Dave Kelsey will follow-up with DEISA and TeraGrid security officers to cooperate on security incidents
• Dave Kelsey will follow-up with DEISA and TeraGrid on AUP alignment
• David Wallom will distribute a spreadsheet to gather resource allocation information in a consistent manner from the infrastructures
• Bob will contact NAREGI to determine if they wish to continue to participate
• Steven & Ruth will organize the next meeting at OGF26