LHCb and ATLAS grid activities

The LHCb particle physics experiment is developing an object orientated software environment known as Gaudi. This framework is intended to apply to all event data processing applications including trigger, simulation, reconstruction and user analysis. To meet these goals, Gaudi has been designed to be easily customisable to different tasks and to allow for the possibility to incorporate components from other frameworks. One particular framework that it will be important to interface with is the Grid, which will provide the necessary coordinated resource sharing in a dynamic and multi-institutional environment in the LHC era. Although Gaudi has been developed in the context of the LHCb experiment, it is experiment independent, as is illustrated by the fact that it has been adopted by other experiments e.g. ATLAS (ATHENA), HARP (Gaudino), GLAST and OPERA. A fundamental requirement of any Grid-based system is the ability to query and access large amounts of distributed scientific data. Using the Gaudi framework, essential services will not only be implemented for LHCb, but will also provide an extremely powerful tool for the exploration of
e-science.

Sub-Project 1: Interfacing Gaudi to grid-based persistent storage

A major feature of the design of Gaudi is the philosophy that a physics algorithm should not act on data objects stored in a persistent data store but instead use transient data. This requires that there is a
set of services that populate the transient data store from persistent data store and vice versa. In addition, it is envisaged that there will be a number of components dedicated to data selection. This will, for example, provide the functionality to the end-user physicist to select particular events or even permit which objects in a datastore are to be processed by the set of algorithms that constitute the event processing application. We have identified the following tasks and deliverables

  • Evaluate and develop LHCb analysis model and study impact on persistent storage technologies
  • Feedback associated LHCb requirements that may influence development of Grid middleware and DataGrid architecture associated with Grid I/O to EDG WP8
  • Develop Gaudi services for Grid enabled I/O
  • Develop Gaudi services for data location
  • Provide software to support storage and legacy data
  • Liaise closely with UK effort in WP5
  • Adapt Gaudi data converters to use any Grid service implementation
  • Evaluate performance of associated software development in distributed environment
  • Exploitation and testing of development software as part of the Grid infrastructure as part of the Mock Data Challenges

Sub-Project 2: A User Interface for Gaudi-Based Applications

The task is to develop an application (GANGA) that allows the end-user physicist
and the MC production manager to manage both production and analysis jobs,
based on the Athena/Gaudi software framework in a Grid environment. Both
Athena/Gaudi applications and GANGA will access Grid services. The GANGA
interface will also serve as a monitor and control interface for running jobs.

The main objectives for this project are:

  • Requirements, architecture, design studies for the complete system to take into account the needs of both LHCb and ATLAS.
  • Gaudi job configuration (including configuration of the Algorithms in the Gaudi application)
  • Event Selection - high-level criteria processing. This interfaces to the book-keeping metadata
  • Job Submission - interfacing to the WP1 middleware via JDL etc.
  • Resource estimation, Resource Brokering
  • Application job monitoring (analysis and production) -interfacing from Gaudi framework to Ganga (counters, histograms, allocations status etc.)
  • GUI - to allow using this functionality by physicists

Sub-Project 3: Monitoring and control of grid applications and recovery from errors.

The studentship will be focussed on developing a system which will allow monitoring and control of grid-based applications. This will build on the generic WP3 middleware and will allow inspection of the history and current state of jobs. The information will be presented to the user in a simple way so that any failed tasks can be quickly identified. The project will also include the development of strategies for performing error recovery and for improving fault-tolerance in the system as a whole. Much of this work will be performed in collaboration
with the LHCb grid group at Rutherford Appleton Laboratory who are contributing to the development of the LHCb Production Monte Carlo system.

Categories: Grid | Particle