---

Computing Technical Design Report

8.2.1 Planning Methodology

The deliverables of the Software & Computing project are determined by the need to prepare a complete and coherent software suite and computing environment for the beginning of ATLAS data taking in 2007. At the same time, the project has to provide continued support for detector commissioning activities and for physics studies.

The continued developments of core as well as of algorithmic code and of the computing environment prompted us to organize a series of so-called Data Challenges, starting in 2001. The scope of the Data Challenges is the test of increasingly complete software and infrastructure at progressively larger scales:

• DC0 in 2001 proved that we could simulate and reconstruct a sizeable number of events (105) with the same software release.
• DC1 in 2002-2003 proved that we could simulate, reconstruct and store in a distributed way ~15M events (still generated with Geant3 and reconstructed in the old FORTRAN-based framework).
• DC2 in 2004-2005 proved that we could use the Grid to generate ~15M events using Geant4, and reconstruct them with the new C++ algorithms in the Athena framework. Later in 2005 the distributed analysis system will be tested on these, and other, data samples.
• DC3 (also called "Computing System Commissioning") in 2006 will test all components of the Computing Model that have not so far been integrated with the mainstream software developments, such as the simulation of the trigger chain and the calibration and alignment procedures (and the simulation of "imperfect" data).

Complementary tests to the Data Challenges are those carried out in the context of the LCG Service Challenges. While the main aim of the Service Challenges is to set up the world-wide infrastructure that will allow the operation of our Computing Model, it is important that our production tools be integrated in those tests, so as to be tested extensively themselves at the same time.

Given these external constraints, it is the task of the Planning Officer, in consultation with the CMB and SPMB, to define the development and release plan for the necessary components. The development plan is then implemented in the Work Breakdown Structure (WBS) [8-1] and the details of the work packages are discussed with the appropriate groups. The WBS is revised approximately once per year.

Progress is followed through regular internal reports; currently the Project Progress Tracking (PPT) system [8-2], developed at CERN for the LHC accelerator and detector construction, is in use. The PPT system is clearly designed for construction projects, therefore we will have to evaluate in the near future how to adapt it (or which other system to use) to monitor continuous software development and optimization activities.

8.2.2 Data Challenge 3: Computing System Commissioning

In preparing for ATLAS physics data in 2007, two complementary planning activities are under way. One of these deals with the use of the offline software and computing system in support of the commissioning of the ATLAS detector sub-systems as described in Section 3.12 , "Use of Offline Software for Detector Commissioning". The other focuses on the commissioning of the computing system itself, the goal being to ensure that it meets minimal functionality, performance and robustness requirements for experiment turn-on.

Detailed planning for the computing system commissioning is under way, with the following major tests being planned:

• Offline Software Chain
• Tier-0 Scaling
• Calibration and Alignment
• Full Trigger Chain including Monitoring
• Data Distribution
• Distributed Analysis
• Distributed Production (Simulation & Re-processing)
• Full TDAQ/Offline Chain.

The strategy is to allow, as much as possible, for these tests to take place in parallel, although some couplings obviously exist and will be factored into the detailed scheduling.

A set of detailed acceptance criteria is being established for each test and these will be incorporated into the recurrent testing and validation procedures described in Section 3.13 to ensure that these criteria continue to be met. They set functionality, technical performance, and physics performance thresholds that need to be met or exceeded. For example, acceptance criteria for the Full Software Chain will include:

• Validation that the output from each stage in the processing chain is readable by the next stage.
• Demonstration of reconstruction from byte-stream files as well as POOL files.
• Non-recoverable error rates.
• Event processing times.
• ESD and AOD file sizes.
• Job start-up time.
• Memory usage.

The TDAQ/Offline full chain test is the culmination of the commissioning process, corresponding to the software and computing systems being ready for ATLAS physics.

This commissioning activity is scheduled to start in autumn 2005, and continue through to ATLAS turn-on.

8.2.3 High-Level Milestones

The list of milestones is intimately connected to the WBS. Here we list only the high-level milestones, i.e. those that, if missed, would have an impact on the ability of the Software & Computing project to satisfy its users.

• July 2005: test the TAG and event-collection infrastructure for data access for
  analysis (important for the Computing Model and the
  Event Data Model).
• July 2005: decision on the future baseline for the Distributed Analysis System.
• September 2005: new version of the Production System fully tested and deployed.
• September 2005: delivery of software and infrastructure for ATLAS Commissioning and
  for Computing System Commissioning; start of integration testing.
• October 2005: review of the implementation of the Event Data Model
  for reconstruction.
• October 2005: Distributed Data Management system ready for operation with LCG
  Service Challenge 3.
• November 2005: start of simulation production for Computing System Commissioning.
• January 2006: production release for Computing System Commissioning and early
  cosmic-ray studies; completion of the implementation of the
  Event Data Model for reconstruction.
• February 2006: start of DC3 (Computing System Commissioning).
• April 2006: integration of ATLAS components with LCG Service Challenge 4.
• July 2006: production release for the main cosmic-ray runs (starting in
  autumn 2006).
• December 2006: production release for early real proton data.