Draft Report of the Muon Reconstruction Selection Panel

[Version 16 Nov 2011]

1. Selection Panel Mandate and Membership

The Muon Reconstruction Selection Panel is charged with selecting the components to form a single baseline standalone muon reconstruction chain to be used at Tier0, data reprocessing and for the processing of simulated data in 2012 onwards. Thus the timescale for making the decision is by the end of 2011 such that the new standalone muon chain may be fully integrated, further tuned for optimal performance and operational in time for the beginning of data taking in 2012. It is possible that components of this chain may need to be rewritten in C++ so as to be fully compatible with the ATLAS software standards and recommendations, in which case this should proceed in parallel during 2012 with the goal to be completed by the end of that year.

The Selection Panel is separate from the muon standalone reconstruction working group, but may call upon their expertise to provide technical information upon which a selection may be based. This is expected to include physics quantities such as efficiencies, fake rates and resolutions as well as technical quantities such as cpu time, memory usage and robustness. These should be under existing LHC luminosity conditions, but also under foreseen higher luminosity and machine background conditions. If need be more qualitative information such as maintainability and conformance with the ATLAS software standards will also be taken into account.

The composition of the Selection Panel is:

David Quarrie (Chair)
Edward Moyse
Alexander Oh
Aleandro Nisati
Ludo Pontecorvo (ex officio)
Fabio Cerutti (ex officio)

2. Summary of Recommendations

3. Introduction and Background

ATLAS management formed the Muon Integration Task Force in June 2010 (See Appendix A), being charged with exploring the technical feasibility of integrating the two muon reconstruction chains (Muonboy and Moore) into a new unified muon reconstruction framework with the ultimate goal of creating a new single reconstruction chain that incorporates the best algorithmic features of the two existing chains. The task force should also propose a path toward integration.

The task force submitted it's final report on 15 Nov 2010, concluding that a unified framework was indeed technically achievable, and identifying several of the components and Application Programming Interfaces (APIs) that would facilitate the possibility of configuring a new reconstruction chain with possibly different implementations of components sharing the same APIs. They implemented some proof of concept APIs and components in support of these conclusions. They furthermore proposed that a new steering framework be implemented to provide more flexibility in configuring the single processing chain.

One of the recommendations in the task force report was that ATLAS and Muon Management should endeavour to provide additional personnel in order to facilitate both the creation of a prototype common chain capable of running with components based upon modified components of the existing Moore and Muonboy chains, but also to allow for the eventual migration of final components to conform to ATLAS software coding and language standards if necessary.

Following the termination of the Task Force, a working group was established, involving both members of the task force, but also new personnel, thus fulfilling this recommendation of the task force. This working group has implemented a prototype based on the following components:

  • A common steering framework, being highly configurable and therefore providing flexibility, allowing for the tuning of different components.
  • Two segment making components, one based on the equivalent Muonboy component, the other on the equivalent Moore component.
  • Two track finding components, one based on the equivalent Muonboy component, the other on the equivalent Moore component.
  • Extensions to the Event Data Model to allow the results of the new processing chain to be made persistent alongside the results of the existing Moore and Muonboy processing chains, thus allowing them to be validated by comparing them against the exiting chains, not only at the end of standalone tracking, but also to be fed to the muon combined tracking chain and then being validated at the point where the reconstructed muons are made available for physics analysis.

The Muon Combined Performance Group has been particularly active in using the eventual availability the results of a single standalone tracking chain to both extend and simplify the combined performance processing chain and the information that's made available for physics analysis.

It is planned that the ATLAS High Level Trigger will also be able to make such optimisations to take advantage of the new chain, but the timescale and validation requirements are different than for the offline processing at Tier0 and reprocessing on the ATLAS GRID sites, which is the initial focus of this work.

Once the new common steering and other components were made available, they were subjected to validation on both simulated and real data samples, and have now reached the stage where they have reached a state of maturity for extensive validation based upon large samples of events of different types, and also the stage where direct comparisons can be made between the different component implementations, leading to the selection of particular components.

ATLAS management therefore formed the Muon Reconstruction Selection Panel in August 2011, with the mandate as described in Section 1 above.

4. Quantities, Samples and Conditions

The Section Panel has made a preliminary assessment of the quantities they would like to see measured on which samples of events and under which conditions. The selection of the baseline configuration will be based on this information.

4.1. Quantities

4.1.1. Quantitative

  • Efficiences as a function of pt, eta and phi
  • Fake rate and Purity both before and after isolation cuts.
  • Resolutions as functions of pt and eta (Z, Z-prime, W?)
  • Performance (CPU, memory, robustness)

4.1.2. Qualitative

  • Maintainabilty
  • Conformance with ATLAS software conventions and tools

4.2. Samples

The samples should be of O(10^5) events. The complete details of the samples used during the evaluation is given here. This section just lists the type of samples and the motivation behind each of them.

4.2.1. Simulated Data

ID Channel Priority Primary Quantities Motivation
1 ttbar 1    
2 Z->mumu 1    
3 Z-prime-> mumu 1    
5 W->munu 2    
6 Exotics 2   High-multiplicity low-pt muons, lepton jets, etc.
7 B-Bbar 2    

4.2.2. High Pile-up Samples

ID Channel Priority Primary Quantities Motivation
1 Z->mumu (mu=40) 1    

4.2.3. Real Data

ID Channel Priority Primary Quantities Motivation
1 Z-skim 1    
2 Muon stream tag&probe 1    
3 J/Psi 1    
4 Cosmics 1   Primarily functionality test

4.3. Conditions

  • 5x10^33 luminosity
  • 1x10^34 luminosity (25ns bunch structure)
  • 3x10^34 luminosity (25ns bunch structure)

5. Standardised Plots and Performance Information

In order to facilitate the comparison between the components, the following plots were produced for all samples, for all component combinations, and also for the baseline Muonboy and Moore processing chains:

  • Efficiency vs. Eta (for combined, tight, medium, loose constraints, corresponding to different stages of the muon combined processing chain [details?])
  • Efficiency vs. Pt (both integrated and differential)
  • Purity vs. Eta
  • Purity vs. Pt
  • Efficiency*Purity vs. Eta
  • Efficiency*Purity vs. Pt
  • Fake Rate vs. Eta
  • Fake Rate vs. Pt
  • CPU time per event
  • Virtual Memory per event
  • Robustness (rate of catastrophic processing failure)

The complete set of plots and performance tables is available here (needs link).

6. Detailed Comparison and Conclusions

Appendix A: Muon Integration Task Force

Charge

With the arrival of collision data we are beginning to learn about the performance of both Moore and Muonboy reconstruction chains in a proton collision environment.

The task force will explore the technical feasibility of integrating the two reconstruction chains into a new unified muon reconstruction framework with the ultimate goal of creating a new single muon reconstruction chain that incorporates the best algorithmic features of the two existing chains. The task force should also propose a path toward integration.

Specific points to be addressed by the Muon Spectrometer Reconstruction Integration Task Force are:

  1. Describe in detail the different components of the Moore and Muonboy suites, including event data model, geometry model, track extrapolation and track fitting tools. Potential advantages/disadvantages of each approach in terms of technical (memory, CPU) and physics performance should be outlined for each component.
  2. Explore the technical feasibility of varying degrees of integration of both reconstruction suites within a new common framework. This includes determining the feasibility of using the same event data model, geometry model and track extrapolation/fitting/material tools.
  3. Identify advantageous features of each algorithm that should be preserved and adopted in the ultimate single reconstruction chain.
  4. Provide a description of the requirements and technical implementation of a unified framework that is consistent with standard ATLAS EDM and services, as well as providing a high degree of modularity for further algorithmic development.
  5. Establish a path toward integration, determine the extent of the work required to bring the two reconstruction chains under the same framework. Provide a phased migration plan with an estimate of the associated time scale and required effort (number of FTEs).
  6. Assess the potential impact of integration on reconstruction clients (combined reconstruction, trigger, validation/monitoring).

A preliminary report (covering 1-2-3) is expected by the end of July 2010 and the final report (covering all points 1 through 6) is expected by mid September 2010. Moreover, a presentation of the work-plan is expected for the Copenhagen ATLAS week.

Membership

  • Members: Ahmimed Ouraou, Niels van Eldik, David Quarrie, Ed Moyse
  • Ex-officio: Software Coordinator (D.Rousseau) Muon PL or Deputy (Ludo or Fabio) 1 representative from ATLAS Management

Summary and Report

-- DavidQuarrie - 16-Nov-2011

Edit | Attach | Watch | Print version | History: r1 | Backlinks | Raw View | WYSIWYG | More topic actions
Topic revision: r1 - 2011-11-16 - DavidQuarrie
 
    • Cern Search Icon Cern Search
    • TWiki Search Icon TWiki Search
    • Google Search Icon Google Search

    Sandbox All webs login

This site is powered by the TWiki collaboration platform Powered by PerlCopyright & 2008-2020 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
or Ideas, requests, problems regarding TWiki? use Discourse or Send feedback