Data Transport RG

Chairs: Pascale Primet, Roger Kosak
Secretary: Neil Chue Hong

Approx 34 people

Pascale reiterated the goals of the group. She then went through the to
do list. Queries were raised about the status of Milestones 2 and 3. The
Extensions to HTTP Milestone data was revised to GGF10.  It was noted
that coordination was still required with the GHPN, RNG and IETF.
Pascale had discussed IETF liason with Cees de Laat. Peter Clarke noted
that Cees could not handle liaison for all groups within GGF and the
group should identify someone who could act as liaison and perhaps
someone from IETF who could attend the Group sessions.

Bill Allcock presented the work done on surveying transport protocols. A
web page had been created listing protocols with descriptions. Some of
these had been started, but none were past draft stage. He suggested
that those not written by the author of the standard should be run past
original authors for approval. 

There was a discussion on the document title and how restrictive it
should be. It was eventually decided that the working title should be
"Reliable Unicast Protocols for Grid Data Transport". Multicast would be
specifically excluded (possibly another document) but message data
transport would be included. The name would be tweaked if necessary to
reflect what's in the final doc. New protocol submissions would be
accepted until September 1st, a draft would come out at October and it
would be submitted after GGF9. A doc, pdf and document template were
available from URL (posted to mailing list)

Reagan Moore presented on "Recommendation for Standard Operations at
Remote Sites". He discussed the overlap between different groups and
their core set of operations. He discussed the drivers for different
groups across the world. Need a minimal set of operations provided
across production data grids. Specify capabilities rather than
implementations.

A question was raised about the linkage with the Storage Management
Group - would the SMG implement things defined by this document.

A new title was proposed "Identification of range of remote operations
in current use". Filesystems were added to the scope of the document.

Reagan was looking for help mapping to implementations e.g. AFS etc.

Igor Mandrichenko gave a brief overview of the status of the
Improvements to GridFTP doc. This had moved to the rechartered
GridFTP-WG

Yves Deneulin gave a presentation on "A Parallel Data Transport Solution
for Grids". Wanted feedback on separation of metadata/data. Description
of how a file is stored (metadata), striping and replication.

This could be used to evaluate different transport protocols.

There was some discussion on this implementation focusing on whether it
was "standard" NFS (it looks like NFS at the client end but is not
elsewhere) and whether this was a file system or a transport (actually a
bit of both).

Pascale finished with a brief overview of "work in progress" and called
for volunteers to take on authorship of the "Transport Scenarios and
Requirements" and "Evaluation of alternative bulk data transport
mechanisms".