DGI-2 Work package 1

Support of Grid users and resource providers

 

| WP 1 | WP 2 | WP 3 | WP 4 | WP 5 | WP 6 |

Support of Grid users and resource providers. Work package 1 (Support) will involve the creation of a central, uniform point of contact that grid users and resource providers can turn to for help in the event of problems or if they require support for the adjustment of resources and applications for the Grid. The Middleware distributed centre of competence also offers a wide range of expertise to support users and resource providers.

The DGUS Portal (D-Grid User Support, dgus.d-grid.de) can be accessed by all D-Grid users that require some form of support. For example, this includes problems with certification, difficulties in registering resources or virtual organisations (VOs) and any other obstacles that occur when using the Grid. As a basic principle, any problems reported are relayed to proficient personnel or groups who then provide solutions.

The form provided on the Portal allows users to submit a description of their problems which is then entered into the database. The first line support team in the Forschungszentrum Karlsruhe then relays these tickets to the responsible party. Each member of the support staff also enters the steps they have taken into the database. Thus, over 600 problems have already been documented in the DGUS database. Over 100 support staff in computing centres throughout Germany are involved in providing solutions to tickets.

The more the DGUS Portal is used and, therefore, the more problem solutions are documented, the faster problems will be solved (or even avoided) in future. It is planned to automate this evaluation during the course of the project.
The Middleware competence centre is responsible for providing consulting to and supporting (level 2 support) grid users as well as resource providers. Support is provided by experts in the participating research centres for the three Middleware program packages UNICORE, gLite and Globus Toolkit. The DGUS system is used as a tool for reporting and tracking problems.

Application-specific problems are also covered in addition to the basic services of the Middleware packages. In particular, this includes support for application-oriented packages such as the "Grid Application Toolkit" (GAT) and the "Simple API for Grid Applications" (SAGA). GAT and SAGA provide a uniform application programming interface (API) for accessing the Grid. Access to the various grid middleware packages occurs via adapters. Thus, it is only necessary for application developers to have a knowledge of the more basic GAT API, rather than the APIs of the individual Grid Middleware packages.

The D-Grid is an amalgamation of a multitude of computing centres. The reference installation represents a computing centre (or "site" for short) of this nature with all supported software components. The support function of this reference installation is to reproduce errors in a defined environment and act as a test platform for new software versions.

Before a prototype site can be developed, it is necessary to establish a computer cluster consisting of multiple computers and a variety of software so that it can be configured and operated automatically. One computer is configured as an "interactive node" and allows administrators and programmers to see inside the cluster.

Because users rely on the grid software to access the individual computers in a grid rather than accessing them directly, computers with user interfaces (UIs) are provided for the various Middleware packages established in the D-Grid (Globus Toolkit, UNICORE, gLite). User interfaces also exist for the data required for working in the Grid. These are SRM/dCache and OGSA-DAI.

The software packages used in the reference installation are also available to the computing centres that are involved in the D-Grid. Software is subject to a continuous process of change in terms of further developments and improvements. As such, the reference installation also has to be continuously updated, although obsolete versions are retained on virtualised hardware.

The update process is carried out twice per year and consists of three cycles: discussion, beta and publication. The current needs of users, operators and software manufacturers are all included here. Details are available on the D-Grid Portal (http://www.d-grid.de).