WP4 Logbook: Frameworks and Visualisation.

Objectives:

• ATLAS validation framework
• Physics framework and tools
• Event Visualisation

ATLAS Validation Framework

Purpose:

Provision of a validation framework for ATLAS developers (including the HLT)

Clients:

ATLAS offline release coordinator, ATLAS analysts, ATLAS developers, ATLAS commissioning, TDAQ

Success:

Adoption of validation framework by significant communities and adoption into the release cycle

Risks:

Failure to provide robustness to the validation framework.
Instability of computing cluster(s) used to run the tests

L1 Deliverables

L2 deliverables

Physics framework and tools

Purpose:

Development of fast simulation tools

Clients:

ATLAS analysts

Success:

Use by ATLAS members to perform physics analysis

Risks:

Full simulation is used exclusively to perform physics studies

We are adopting a cyclical approach, which is driven by the offline release schedule This is currently known through release 14.0.

The L2 objectives for each L1 are:

• Ensure that the fast simulation is running in the ATLAS environment, and update the documentation to reflect changes when these have been found to be necessary. Maintain a testing program.
• Run comparisons between fast and full simulation using the Comparator developed as part of the ATLAS-UK EScience project. This will be done over a range of variables, data sets and selection cuts. Run comparisons with data when this becomes available.
• Model the effect of detector upgrades in the fast simulation.
• Re-factoring and generalisation of AOD example code
• Develop tests for physics analysis code. Integrate these tests into the validation framework. Follow up the test results.

L1 Deliverables

Event Visualisation

Purpose:

ATLAS event display

Clients:

ATLAS analysts, ATLAS developers, ATLAS commissioning, TDAQ

Success:

Adoption of ATLANTIS by client communities

Risks:

Demands from various clients overwhelming the Atlantis effort

L1 Deliverables

Commentary

04Q3

The first deliverable for the validation activity is due in one month, being a requirement capture and design document for the broader validation service. Both aspects exist in draft, and RTT represents a major component already being deployed. This RTT prototype has consumed much of the effort, and is now providing a service for target communities. The detailed level-2 deliverables can be seen at http://www.hep.ucl.ac.uk/atlas/EScience/Planning/EScValidation.html. An important issue is the robustness of the product, and we are attempting to identify other hardware resources that can run the testing as a fail-over option. In the medium term, the testing will be done on Grid resources.

Analysis Framework and Tools: The Artemis activity now uses the AOD framework as decided upon after the Artemis workshop in April 2004. This framework is also receiving support from BNL and CERN. We have aimed our efforts (which are for the moment reduced to allow more effort in the validation area) at producing real demonstrators in physics groups, both to encourage adoption and to learn of the real features required. The primary demonstration is in the UK-lead B-physics area (Catmore, Smizanska), where example code and documentation has been produced. http://msmizans.home.cern.ch/msmizans/production/phys_analysesdc2.html The remit has been broadened to include the first use of the Grid for analyst-users, which has also been extensively documented. http://msmizans.home.cern.ch/msmizans/production/dc2/grid3.html The first deployments in both cases are workable but unsatisfactory, and a strategy and design review is planned over the next few months.

Visualisation: Most of the effort over the last six months has been focused on the support of the ATLAS Combined Testbeam (CTB). This involved developing an automatic procedure for accessing the CTB geometry from athena and implementing it into atlantis. A server-client communication between athena and atlantis was also developed, based on the standard xmlrpc protocol, for using atlantis as an online event display at the CTB control room. Atlantis is being used successfully in the CTB and there are several cases where it helped both in spotting problems in the hardware and in fixing bugs in the detector reconstruction software. Significant effort has also been devoted to user support and developers documentation, as well as in testing and following up updates in the rest of the ATLAS software.

04Q4 and 05Q1

The two L1 requirements for the period 2004-Q4 and 2005-Q1 were the requirements document, and the the prototype release. Both were delivered on time. During the prototype development work, a number of of unanticipated situations arose and were dealt with. The most important of these was the dropping of Linux Red Hat 7, which was replaced by Linux SLC3, and the necessity to run from ATLAS software distribution kits.

The transition to SLC3 was smooth. Prior to the switch over of the UCL farm, where the nightly tests are run, the validation software was run nightly in on a desktop machine with a cut down work load that was chosen to have at least one job from each type found in the full work load. The only feature required specifically by the validation software was a build of Python v2.3 on the SLC3 platform.

Until the end of 2004, the official ATLAS stance was that the release on AFS was the official release. Due to kit patching, the distributed software could, in fact, deviate from the official releases. For this reason it became clear that running from kits is necessary validator functionality. A further reason is that the kit itself needs validation as it is being constructed. A battery of small tests are run using the KitValidation package, but the full power of the validator is also required.

While gaining experience with the prototype it became clear that the validator, was very highly centralised. Complex tasks were carried out by specially written scripts. The flexibility of the software lay in its modularity, not in its control structures.

To remedy this, many variables were identified as input parameters, and were stripped out of the code and placed in configuration files, thus converting specialized functions to generic scripts. These configuration files are stored in the package they refer to, and a discovery mechanism to find which packages provide such information was set in place/ This greatly improved the validator flexibility as maintenance of the configuration files is now the responsibility of those who have the necessary expertise.

The ability to run user supplied scripts to carry out specialised tasks not handled by the validator is an important feature. Information which allows user supplied scripts is possible through the configuration files. The validator does not merely call the scripts but also passes information to the client script, allowing a two-way interaction between the validator framework and the client scripts.

The amount of information provided by the RTT can be bewildering. Jobs can be run form different job groups, with very run time needs, tests, and results files. The same jobs can be run under different circumstances including different releases, debug or optimised builds or from kits. Our efforts in making the information available in an absorbable manner are aided by user feedback from the prototype.

The validator software was reviewed under the general ATLAS Software Infrastructure review in February 2005 (see )

WP4.2 Graphics

The L1 milestones of using the Atlantis event display used in the 2004 Combined test beam (1/10/04) and documenting the use of standard protocols for online data access and use as a web service (1/04/05) were both met.

Many new features have been introduced to the ATLANTIS event display program. For the calorimetery, energy colour maps, log energy scale, and lego plots were provided. Missing Et is now displayed. Events, which were transferred through the network can be saved as XML files. A user's configuration can be saved, and used it at the following start-up. The GUI has been improved.

Online data access and Atlantis-Athena communication has been implemented using XMLRPC protocol as described at

http://www.hep.ucl.ac.uk/atlas/EScience/Documents/Visualisation/jiveXML-atlantis-communication_design.pdf

Data can be send from ATLANTIS to ATHENA allowing interactive ATHENA "driven" from ATLANTIS. Execution of ATHENA algorithms is now possible from ATLANTIS

Work to ensure ATLANTIS will be usable for ATLAS commissioning has started, and includes a first look at simulated cosmic data for inner detector, and discussions with Tile commissioning team.

A number of ATLANTIS tutorials have been given, and the next one is scheduled for the 27th May, during the ATLAS software week. We have Atlantis developer's guide, which can be seen at :

https://uimon.cern.ch/twiki/bin/view/Atlas/DetDescAndGraphicsReview/atlantis_developers_guide.pdf

and the Atlantis start-up guide, which is viewable at http://cern.ch/atlantis/docs/AtlantisStartupGuide-v12.pdf

ATLANTIS was reviewed in the ATLAS software graphics review. The proceedings of the review can be found at:

https://uimon.cern.ch/twiki/bin/view/Atlas/DetDescAndGraphicsReview

WP4.3 Physics Analysis and Tools

After discussions with the ATLAS ESRAT (Event Selection, Reconstruction and Analysis Tools) co-ordinator, it became clear that there was a need to develop tools beyond the idea of a simple event data model, and accompanying helper functions. The work at UCL on the ATLANTIS event display, in particular interaction with ATHENA provided an opportunity to provide a vital set of tools that allow physicists to visualise the reconstructed entities, in order to develop, debug and tune reconstruction and physics algorithms. Early work to exploit the XML-RPC transport mechanism put in place as part of WP4.2 were presented at the Tucson Physics Analysis Tools workshop in April 2005.

Other work being carried out in the context of the ATLAS AOD is the development of event views. An event view provides a coherent, exhaustive and mutually exclusive handling of reconstruction entities.

05Q2

Over the last quarter there has been a growing usage of the validator process. By the end of June, 12 job groups being processed for two build branches.

As part of the co-ordinated ATLAS offline testing effort, the RTT post processing tools into separate python modules which are installed as part of the offline software build process. The post processing tools now conform to a common interface. The combination of these changes allow RTT tools to be freely used by other testing framworks, and brings a greater degree of coherency to the RTT itself.

Down time on the UCL cluster used to run the validation jobs underlined two vulnerabilies of the system: jobs were not run for the period the cluster was down, and for some time afterwards the system suffered instability. The second effect was to show that more extended quality control was required. The backlog of untested - because tests were impossible during the doen time - made working conditions very difficult. The second problem has been cured by instituting a pipeline of tests of increasing completeness. When the pipeline is full, code development must stop until the backlog is cleared. This means the production system is protected, and bugs are in new code are identified and fixed far more rapidly than before.

Two approaches are being followed to solve the problem of vulnerablity to hardware failure or unavailablity: the first if to find a back-up batch system which would run the validation framework. The other approach being investigated is to run the validation jobs on the grid.

Work has been put into the web pages to improve usablity and clarity. Basic monitoring scripts have been implemented.

WP4.2 Graphics

The work in this quarter involved:

• introduce or improve features that were requested by users (e.g. missing ET, Regions of Interest, Lego plot)
• prepare and deliver the 4th atlantis tutorial; the tutorial was once again over-subscribed and the feedback from the participants was very positive
• improvements in the atlantis-interactive athena communication (e.g. secure communication via ssh, asynchronous calling)
• support of sub-detector commissioning activities (e.g overlaying of events, discussions/explanations/demos)

WP4.3 Physics Analysis and Tools The physics analysis effort is concentrating developing and extending existing areas of expertise at UCL, namely fast simulation (Atlfast) and interactive analysis using ATLANTIS. Creation to export AOD objects for Atlfast has been updated and Atlfast track objects have been added. The Atlfast jobs have been updated to use the DC2 data format. Test jobs have been added to the RTT. Modifications to Atlfast to allow the use of the WP3 comparator aer underway. Work on providing fast simulation for monopoles has started.

05Q3

WP4.1 Software Validation

The validation software has demonstrated a much higher level of stability this quarter due to the efforts made during 05Q2 to improve our self-testing protocals. The RunTimeTester still experiences residual instability due partly to failure in systems on which it depends, and, at a very low rate, due to a number of reasons apparantly due to the cluster and batch queues on which it runs. We are working to identify the causes more precisely. Work on the installation of a parallel backup version of the validation package has been started at Lancaster, to cover periods when the UCL cluster is unavailable. The number of client offline packages has ridden to 19. The success rate for running test jobs varies, put was high (approx. 80\%) at times towards the end of the quarter. The number of user provided post job tests remains low. Effort has been put into providing RTT monitoring. This enables better control of the RTT for the development team, and a better understanding of the state of software testing in ATLAS for managers, in particular the SIT co-ordinator. The RTT continues to work to improve monitoring. Internal reorganisation of how test results are stored has been performed. This was necessary to accomodate new clients who report their results via linked web pages. To improve throughput on the cluster in the 24 hour testing period (the tests are run once every 24 hours after the nightly builds have completed) have motivated us to use threading to increase parallelism in the system. As a results of the change in the operating system (CERN moved to SLC) we lost the ability to run the RTT interactively, which is vital to allow package testers to develop their tests. The reasons were identified by the RTT team - the problems included incorrect python versions for ATLAS, and incorect PyXML installation by the LCG - and interactive running has been restablished. As a further aid to testers we have set up an interactive web based configuration file tester to further aid developers to integrate their tests into the RTT.

WP4.2 Graphics

The work during this quarter has focus on the following areas:

• Prototype and demonstrate an example for athena algorithm execution. The algorithm chosen is secondaring vertexing, which ties in well with the other ATLAS eScience activities in WP2. A demonstration is being prepared that will take place during the Physics Analysis Tools meeting at the ATLAS software workshop.
• New features and improvements, following user feedback and requests. Examples:
  • The tool which accesses the event data in athena and produces the XML files, has been turned into an AlgTool, so now it is possible to run athena over a large number of events and produce XML files only for those events of interest.
  • Display of secondary vertices, track residuals, missing ET in eta-phi projection, jets in the lego plot, LVL1 Trigger tower information, AOD data types.

WP4.3 Physics Analysis and Tools

Physics analysis work this quarter has centred around developing and maintaining the ATLAS fast simulation package, Atlfast. A new scheme of jobOptions has been introduced to afford easier access to new users, thus extending the user base within the collaboration. The electron and muon classes have been adapted to work effectively with the WP3 comparator tool, enabling accurate future calibrations to real data and/or full simulation. New classes have been designed and integrated to accommodate monopole production. Progress has also been made towards graphical analysis with Atlantis, the ATLAS event display. JiveXML has been successfully run on DC2 data and extra functionality has been introduced to allow AOD objects to be displayed in Atlantis.

05Q4

WP4.1 Software Validation

Increased functionality

• Flexible timeouts
• improved error reporting
• Machinery for running short jobs for kits
• Performance enhancements
• Improved CMT usage
• Timing functions
• Improved use of code to call shell commands
• Configuration test file location utlity
• Improved RTT self monitoring and run summary information
• Improved presentation page layout
• Improved error recovery
• Simple tools to test afs connection
• Improved logging

User Control Improvements

• e-mailing on request
• numbered release only jobs

Services

• web based xml file validator

New Job Groups

• Physics AnalysisTools
• G4AtlasAppsRelease
• Analysis examples
• BPhysValidation
• MuonBoy
• MuonBoyPerformance

Self Testing

• Phase1 RTT unit/regression tests
• Phase2 RTT unit/regression tests started

WP4.2 Graphics

There have been three major activities during this quarter:

• Interactive execution of algorithms: The prototype work was completed and the whole mechanism became more transparent and easier to use for the general user.
• Displaying of AODs: prototype versions of atlantis and JiveXML have been given to the Physics Analysis Tools group for feedback.
• Commissioning: there has been close interaction and collaboration with the Inner Detector commissioning team, and the requested features have started to appear in atlantis. These include, visualisation of geometry details of the Pixel/SCT detector elements and TRT sectors visualisation of multiple collections of a given Datatype, overlaying events, providing more information about clusters/spacepoints etc.

WP4.3 Physics Analysis and Tools

• the Atlfast Event Data Model was extended to allow for AOD-style navigation
• work has been carried out to allow Atlfast to work with the Comparator (WP3). Many parameters of the program, which have been protected from easy modification in the interest of program stability have been exposed to allow the Comparator to carry out its optimisations for photons, electrons and muons.

06Q1

WP4.1 Software Validation

New Functionality

Major Changes

Improvements

• web pages
• logging

Refactoring

• streamlining logic
• use of clearly identifiable factory methods
• modifications to facitlitate testing

New Job Groups:

• tauRec
• TriggerReleaseReco
• TriggerReleaseBS
• InDetRTT

User Control additions

• command line flags for athena
• improved packages vetos

Self Testing

• improve regression point evaluation
• improve portability
• enable the RTT to call its own test suite (allow recursive calls)

WP4.2 Graphics

Intense development following requests and feedback from users has continued. The most significant items in this period have been:

• Further improvements in the XML parser. This work has been vital since with the intense development in atlantis and JiveXML, users with older versions of atlantis would suffer when producing XML files with new information as in the older versions of atlantis XML fragments that were not recognised would cause atlantis to crash. The new parser produces warnings and in some cases has the intelligence to tell the users to upgrade to a more recent version of atlantis.
• Reading in and displaying multiple collections of a given DataType (e.g. jets or tracks). This was a request from many communities: Inner Detector offline, commissioning, physics analysis (for jets). We developed a uniform way of handling the problem of displaying multiple collections that could apply to all DataTypes.
• Geometry changes. A simplified, much tidier and economical description of the geometry, especially the cells of the various calorimeters. Also the Detector Elements of the Silicon Trackers (used heavily in InDet commissioning) and the TRT Sectors.

WP4.3 Physics Analysis and Tools

The work of the ATLAS Atlfast Task Force (ATF) started this quarter. The goal of the task force is to ensure that Atlfast provides is adequate for performing physics analysis. The task force is composed mainly of physicists concerned with detector performance, simuulation and reconstruction, together with the Atlfast team.

The work is being carried out in two phases. In the first phase, the parameterisations used in Atlfast are being examined and, where necessary, updated to reflect changes in detector layout and reconstruction since those developed almost ten years ago.

In particular, changes have been made to:

• labelling and tagging of hadronic taus
• missing ET calculation, for which two alternative techniques were introduced to adjust for energy not considered due to Atlfast's current particle isolation routines
• the scalar sum of ET was introduced for easier comparison with full simulation.

06Q2

WP4.1 Software Validation

WP4.2 Graphics

Work continued on supporting the InDet commissioning, interactive vertexing, improving the displaying of AOD types, improving the functionality of the GUI and the configuration of atlantis, and on training. Specifically:

• We gave a half day atlantis tutorial at the end of the Trigger and Physics week in May and participated in the Physics Analysis Tools tutorial during the Software week in April. Both tutorials had full attendance.
• The displaying of all AOD types in atlantis has been completed. During the tutorials, some more feedback was received by the physics analysis tools experts and the physics community and work is ongoing accordingly.
• Interactive vertexing was improved so that tracks from any track collections can be passed from atlantis into the athena vertexing tool (before it was assumed that these tracks are of a certain type). This work became necessary after the changes where multiple collections are possible to display in atlantis.

WP4.3 Physics Analysis and Tools

06Q3

WP4.1 Software Validation

WP4.2 Graphics

The main activities in this quarter included:

• Major restructuring/tidying up of JiveXML. The monolithic JiveXML was split into nine packages. The core JiveXML package now only depends on a few core packages and has moved to the AtlasEvent project. The other packages correspond to various "sub-systems", e.g. InDet, Calo, Muon etc. Apart from a general tidying up, which was long overdue, the advantage of the new structure is that there will be more developer's releases where you can generate XML files with data from some sub-systems. Also, the sub-system experts can be more directly involved in updating the JiveXML packages following changes in the EDM etc.
• Support of the ATLAS sub-detector commissioning, by adding features according to the requests from the different communities.
• New command processor. This allows to loop continuously over different events and is a very useful feature for demonstrations, public outreach etc. The events can also be accessed through a URL, and the idea is to use this feature for transmitting cosmic (and ultimately real) events through the CERN television system but also via the web around the world.

07Q1

WP4.1 Software Validation

WP4.2 Graphics

The main activities in this quarter included:

• Integration to the online monitoring framework at the ATLAS Control Room.
• Support of the ATLAS sub-detector commissioning, by adding features according to the requests from the different communities. In particular, work with the TileCal and InDet commissioning teams.
• Updates on documentation, web pages, and general tidy-up. Preparation for an Atlantis tutorial in beginning of May.
• New calorimeter features, special views for the endcaps. Additional features for displaying AOD objects.

Meetings & Papers

Effort Delivered

Effort funded by ATLAS E-Science

E. Nzuobontane (UCL)

Quarter	Validation framework	Analysis framework	Visualisation	Total	Grand total
04Q2	0.0	0.0	0.0	0.0
04Q3	1.0	0.0	0.0	1.0
04Q4	1.0	0.0	0.0	1.0
05Q1	1.0	0.0	0.0	1.0
05Q2	1.0	0.0	0.0	1.0
05Q3	1.0	0.0	0.0	1.0

Z. Maxa (UCL)

Quarter	Validation framework	Analysis framework	Visualisation	Total	Grand total
04Q2	0.0	0.0	1.0	1.0
04Q3	0.0	0.0	1.0	1.0
04Q4	0.0	0.0	1.0	1.0
05Q1	0.0	0.0	1.0	1.0
05Q2	0.0	0.0	1.0	1.0
05Q3	0.0	0.0	1.0	1.0

B Simmons (UCL)

Quarter	Validation framework	Analysis framework	Visualisation	Total	Grand total
04Q2	0.8	0.2	0.0	1.0
04Q3	1.0	0.0	0.0	1.0
04Q4	1.0	0.0	0.0	1.0
05Q1	1.0	0.0	0.0	1.0
05Q2	1.0	0.0	0.0	1.0
05Q3	1.0	0.0	0.0	1.0

Qiang Lu (Birmingham)

Quarter	Validation framework	Analysis framework	Visualisation	Total	Grand total
04Q2	0.0	0.0	0.0	0.0
04Q3	0.0	0.0	0.0	0.0
04Q4	0.0	0.0	0.0	0.0
05Q1	0.0	0.0	0.7	0.7
05Q2	0.0	0.0	1.0	1.0
05Q3	0.0	0.0	1.0	1.0

Effort not funded by ATLAS E-Science

Quarter	Validation framework	Analysis framework	Visualisation	Total	Grand total
04Q2	0.5(PS)	0.5(PS)	1.3(JD, NK)	2.3
04Q3	1.0(PS)	0.0	1.3(JD, NK)	2.3
04Q4	1.0(PS)	0.15(JC)	0.45(NK, JT)	1.6
05Q1	1.0(PS)	0.5(SD)	0.45(NK, JT)	1.95
05Q2	1.0(PS)	0.5(SD)	0.45(NK, JT)	1.95
05Q3	1.0(PS)	0.5(SD)	0.45(NK, JT)	1.95

PS	Peter Sherwood
NK	Nikos Konstantinidis
JD	Janice Drohan
JC	James Catmore
JT	Jurgen Thomas
SD	Simon Dean