Datagrid WP1 status report



Status report of datagrid WP1 (Workload Management)
as of February 8th, 2001

Francesco Prelz - INFN Milano
Massimo Sgaravatto - INFN Padova

Summary:
I. Activities to provide D1.1 (report on current technology)
II. Activities towards D1.2 (Architecture definition). Interaction with the ATF.
III. Requirement collection activities
IV. Activities towards M1 and D1.3 (prototype at Project Month 9)

---------

I. Activities to provide D1.1 (report on current technology)
The report on current workload management technology (due for project month 3) will include an assessment of the software developed by the Condor and the Globus projects, drawing on the experience gathered within the INFN Condor group and the INFN-GRID Globus evaluation activity.
We are also planning to include a short review of current research in the field of scheduling, since we are also discussing some longer-term approaches, such as the economic scheduling model.

---------

II. Activities towards D1.2 (Architecture definition). Interaction with the ATF (WP1 ATF rep: Francesco Giacomini).

A block diagram of the currently foreseen components of WP1, along with a sketch of the dataflow and functional interactions among the various blocks (and with other Datagrid WPs) can be found at http://www.mi.infn.it/~prelz/grid/ This diagram is constantly updated as more details are clarified and is the reference basis for discussion within the WP. This is obviously not an "architecture", but reflects the interface and/or encapsulation changes that are negotiated within the project Architecture Task Force. Given the role of the ATF and the fact that the application requirements will evolve as more familiarity with the data grid model is acquired, we are currently not following a classic top-down requirement collection, design and implementation scheme. We are rather trying to rapidly progress towards a working prototype with the fundamental functionality (see section IV). The architecture document at month 6 will reflect the ATF vision, will possibly finalize the contents and boundaries of the WP and the interaction/interfaces
with other WPs, and will describe the short and long term technical choices
that are made.

---------

III. Requirement collection activities
The definition of the Job Description Language (JDL) that is used to specify a computing activity for submission to the grid clearly needs an early and active interaction with the application WPs. Instead of going through a requirement collection cycle we decided to prepare a document (it can currently be accessed from http://www.mi.infn.it/~prelz/grid/classad-howto.pdf) that formally describes the syntax and semantics of a "prototype" JDL, so that applications can see whether it is capable of describing their typical or foreseeable grid computing activities. We are ready to collect feedback on this prototype language so that a suitably extended design can be made These comments should be sent to francesco.prelz@mi.infn.it and massimo.sgaravatto@pd.infn.it so that we can post them to the WP mailing list.

---------

IV. Activities towards M1 and D1.3 (prototype at Project Month 9)
After discussions within the ATF and the WP, the contents of the PM9 prototype we agreed upon are now reasonably well defined. Here's a summary:

The University of Catania, INFN-Pisa and INFN-Torino groups are working on information gathering and scheduling models towards the Project Month 9 prototype implementation of the core "resource broker" engine.
The Datamat group is following the specification and development of the JDL, and is starting to design the prototype command-line user interface.
The CESNET group is modeling the user view of the job lifecycle, and identifying the optimal status and monitoring information collection and gathering locations. The format and distribution of the monitoring data will reflect the choices made by WP3 (monitoring). A communication channel with WP3 needs to be established.
The Grid submission services will be obtained from the Globus software in the prototype implementation, so WP1 is remaining active in following the developments of the Globus GRAM, the Condor-G services and other submission services. This also defines (following feedback from the ATF) the responsibility boundary with WP4 (Fabric Management) in the job submission service area.