xrootd discussion - ATLAS is considering this, STAR and SLAC brining experience for this. How does it implement an SRM functionality?
How will OSG plan to use an SRM/dCache SE? Trash/Deployment on 15 sites.
Miron: we need an open discussion on what we need.
Ruth - drawing xrootd stack.
root integration with dcache - there is a special class tdcache.
xrootd metrics that would be hard to beat - due to its namespace architecture, eg., number files to open /time. block access is desired.
How do they compare in terms of:
local network faults?
faults in a file
Ruth: we need to worry about capabilities not the technology, and the needs of current and future stakeholders.
Then applications must be capable of discovering which protocol to use, given a logical file name.
Timor: glite knows how to open files in any of these types of protocols, e.g.: xrootd, dpm, dcache, etc.
Miron: we need tools and libraries to help smaller VOs to access these protocols.
Issue of supporting two SE's on a site: advertising the resource, how applications connect to them, which CPUs see them, etc.
Note - there is at least one stakeholder in OSG that requsts OSG to consider xrootd as a supported storage service included in VDT, and would asked that the executive team would consider its priority and effort available, milestones of stakesholders, for its evaluation.
Next issue: ACLs and opening files, re-using UIDs, and mapping UIDs to GIDs, and identity management issues. Is the DN a unique key?