identifiers in grids

[Reagan Moore]

The question of identifiers can be subsumed within the larger 
challenge of maintaining consistent naming conventions across 
federations of collections.

Data grids use a logical name space to define global identifiers. 
Each type of physical resource is typically named by a separate 
logical name space:
- files, named within a replica catalog or SRB MCAT catalog
- physical resources, within a grid metadata directory service, or SRB MCAT
- users, named within a certificate authority
- access controls, managed by a community authorization server, or SRB MCAT

Each grid service corresponds to mapping of distributed state 
information to the appropriate logical name space managed in the 
appropriate registry.

Management of consistency across grid services corresponds to 
constraints that are imposed on the mappings.

A simple example is management of access controls on data that is 
aggregated into a container and replicated across multiple systems 
(including compound resources).  In the SRB environment, the 
following distributed state information is managed across multiple 
services for this capability:
- mapping is established between user identity and the SRB name for 
the user through an authentication service
- mapping is established between SRB user name and access control 
lists for each digital entity registered into a SRB collection
- mapping is established between logical file name and a container 
that holds the file
- mapping is established between the container and the physical 
resource that has the desired replica of the container
- mapping is maintained of the presence of the replica on the disk 
front end to the back end tape within a compound resource

These mappings require additional state information for enforcing 
locks on writes, and synchronization across replica sites after 
writes complete.

Access control to the physical digital entity requires the exercise 
of all of these mappings in the correct order.

The issue of identifiers is intricately intertwined with the mappings 
that are imposed on the identifiers.  The infrastructure that manages 
the mappings can also be applied to create digital libraries (mapping 
of descriptive metadata to the logical name space), and persistent 
archives (mapping of authenticity information to the logical name 
space).

Reagan Moore