OGSA Data Session Notes
1 February 2007, 11:00-12:30 EST
Dave Berry Leading

EMS Scenarios

Andreas: When JSDL specifies a protocol, you must use that protocol.

Do we need a scenario where the executable accesses the data directly, effectively using its remote name and having it mounted as part of the local file system.

If BES takes an RNS name, does it need the intelligence to determine where the object
is and arrange for efficient access (e.g., by staging data).

Suggestion to add a scenario based upon taking an RNS name to denote the data to be accessed.  And then let the BES determne how to best support that access (e.g., direct access, stage with an approprite protocol (DMI, Grid FTP)).


It was suggested that the scenario description start at a higher level.  The current scenarios seem to be at an implementation leve

The scenarios presented are best described as "concrete scenarios" rather than as "conceptual scenarios".

Scenario 3: "DTS" -> "DMI"

There was considerable conversatation that leads to the conclusion that these scenarios need to be carefully written to make it clear that they are illustrative.  One particular case is the first scenario which is bound to ByteIO - it should make it clear that one could plug inother access mechanisms such as GridFTO in its place.

It is suggested to fit storage issues into these scenarios.=