The ATLAS Trigger Configuration System Design and Commissioning A.dos Anjos, P.Bell, D.Berge, J.Haller, S.Head, T.Kohno, S.Li, T.McMahon, M.Nozicka, H.v.d

The ATLAS TriggerConfiguration System

Design and CommissioningA.dos Anjos, P.Bell, D.Berge, J.Haller, S.Head, T.Kohno, S.Li, T.McMahon,

M.Nozicka, H.v.d. Schmitt, J.Stelzer, T.Wengler, W.Wiedenmann

Thursday, September 6 The ATLAS Trigger Configuration System 2

Outline

Trigger designConfigurable components

Configuration system requirementsData taking, MC production, offline analysis

Design and implementation

Commissioning status

Conclusions


Trigger Design

Fast, highly selective, efficient

40 MHz bunch crossing

100-200 Hz storage rate

~40ms

~4sec

~3% of detector

introduction to the ATLAS trigger in ATLAS HLTsteering by S. George (Mo 16:50 Online Computing)


Level 1 Trigger

Pre-processor

ClusterProcessor

Jet/EnergyProcessor

End-cap MuonTrigger (TGC)

Barrel MuonTrigger (RPC)

Muon-CTP-Interface (MuCTPI)

Central Trigger Processor (CTP)

LTP

BusyTTC

Detector Front-Ends/Read-out

LTP

BusyTTC

Muon DetectorsCalorimeter Detectors

Common Merger Modules

Trigger objects: Muons, EM and hadronic clusters, jets, total and missing ET

CTP

• Mapping of 480 hardware signals onto 160 CTP internal signals these signals encode the object multiplicities• Thresholds of trigger objects multiple thresholds for different object types• Item definitions: logic, prescale and veto rates maximum 256 trigger items• Random trigger rates, trigger on bunches or bunch groups

Configurable information on CTP:


High Level Trigger (HLT)

Concept of trigger lines (chains)Chain: ordered list of trigger conditions (multiplicities of HLT trigger elements) to be evaluated in sequence)Description how algorithms produce trigger elements (example: L1EM3 ClusterFinder&Hypo L2_e5cl)Collection of chains (with prescale and forced-accept rates) HLT menu (see Teresa’s talk)

HLT algorithms configured through parametersSet via python (used in ATLAS as high level scripting language)

more details in talks about ATLAS HLT steeringby S. George (Mo 16:50 Online Computing)and about Trigger Reconstruction Algorithmsby T. Fonseca Martin (Mo 17:55 Online Computing)

signature (e j)

sequence (e) [EM “e-FEX, e-Hypo” e]

sequence (j) [JET “j-FEX, j-Hypo” j]

signature (e’ j’)

Chain (EJ-L2) input = “EMJET”

Chain (EJ-EF) input = “EJ-L2”

Lvl1 Trigger Item EMJET

y/n

y/n

y/n

y/n

L2EF

HLT Chain

Configurable information:• Chain definitions: logic (trigger conditions, algorithms), prescale and forced-accept rates maximum 8192 chains per menu• Algorithm parameters• Data streams, monitoring groups

Ste

erin

g, s

ee

Sim

on’s

talk


Design Requirements for the Configuration System

Complete and consistent configuration of the ATLAS triggerOnline software and hardware for data takingTrigger simulation software in Monte Carlo production jobsConfiguration information provided to the user to perform trigger aware analyses and trigger studies

Flexible and fast configuration changes during data-taking to react to different beam and detector conditions

History of configurations for the purpose of understanding and reproduction of the trigger behavior


Components of the Configuration System

Relational database stores trigger configuration (TriggerDB)

Trigger configuration via a single keyOffline reproducibilityTrigger historySchema reflects trigger design

Tool for database browsing and manipulation (TriggerTool)

Flexible and fast changes of the trigger during data takingTrigger experts, shift crew, offline analyst

Software clients to directly access the TriggerDB for

Data taking, simulation, and distribution of configuration data (conditions database)

TriggerDB

simulation production

shift crewoffline user expert

TriggerTool

Relational Access Layer

conditions database

datataking

Level 1 Menu + PrescalesHLT Menu + PrescalesAlgorithm parametersRelease version

TriggerDB Schema


Operation of the Configuration System – Preparation

Level 1 TriggerMenu: stable in time, small changes to thresholds and trigger items by hand using TriggerTool

Trigger-Menu-Compiler creates image for Level 1 hardwarePrescales: adjusted by shifter to match the luminosity optimize bandwidth usage

High Level TriggerPrepare and validate trigger menu for data takingPopulate the TriggerDB with the HLT configuration information

Menu, algorithm parameters, prescale ratesCheck consistency with Level 1 configuration

Configuration alias for shifterLogical names (‘PHYSICS’, ‘COSMICS’, ‘CALIBRATION’) pointing to current valid configurations

ATLAS Trigger community responsible for the development and testing of the trigger algorithms to achieve the ATLAS physics goals


TriggerDB during Data Taking

Shifter to chose trigger configuration alias before CONFIG transitionConfiguration key written to online configuration database to be picked up by Level 1 CTP controller and HLT processes

At CONFIG: CTP controller loads image from TriggerDB into the CTP hardwareHLT processes load configuration from TriggerDB into memory and initialize themselves

At START run: partial configuration information is written into the ATLAS conditions database (COOL) as run-wise trigger configuration data

While RUNNING:Level 1 prescales can change written to COOL Trigger Panel in the ATLAS Run Control interface

see talk about ATLAS Online Configuration Database by I. Soloviev

(Wed 15:05 Online Computing)


Trigger Result

Run-wise configuration data to interpret trigger decision (conditions database)

Event wise trigger decision encoded in the bytestream

Maps: trigger names to bit position and chain counterAllows for access to the trigger decision using trigger names

Information about trigger definition at each step of chain-processingTo rebuild the HLT menu and access the trigger objects by name

Prescale, LVL1 veto and HLT forced accept rates and trigger chains

Level 1 trigger: Acceptance flags for the up to 256 active trigger items before and after the application of prescale and veto 3 x 256 bitsHigh level trigger: Acceptance flags for each chain before and after the application of prescale and forced-accept

Chains are identified by a short integer (chain counter)Index of last successfully processed step for each chainTrigger objects e.g. hadronic clusters, muon tracks

Information to link these HLT trigger objects to the Level 1 trigger objects (trigger studies)


ES

D 1

00M

B/s

AO

D 2

0MB

/s

ESDAOD

TAG files/DB

Flow of Configuration DataMore details in talk about ATLAS Databases

by A. Vaniachine (Wed 14:40 Distributed data analysis) and about ATLAS Tag DB

by F. Viegas (Wed 14:40 Software Components)

L1Result

to

Tier0

express calib

Tier

1

tran

sfer

Tier 0

Prompt Reconstruction

Express Reconstruction,

calibration

Tier 1 Reprocessin

gTier 2MC

production

ConditionsDatabase

Trig

ger

Men

us

into

Con

ditio

ns D

B

ConditionsDatabase

TriggerDB

DbProxy

LVL2 Result

EFResult

RODs

Front-end

LVL2

Subfarm Input

EF EF EF

Event Builder

LVL1/CTP

Subfarm Output

EF

Trigger ResultTrigger Objects

1. TriggerDB to configure trigger for data taking

2. Configuration data to COOL

3. Trigger result in each event

4. Shipped to reconstruct-ion sites

5. ESD, AOD, TAG for trigger aware analysis TriggerDB

Replication

http://indico.cern.ch/getFile.py/online/ringberg_jun06.ppt








Trigger Information for trigger studies and physics analysis

Configuration Data in Offline Analysis

TriggerDBAll configuration data

online DB (COOL)

• Configuration: Lvl1 items and HLT chains (name, version), prescale-, Lvl1 veto- and HLT pass-through rates• Trigger result: pass or fail? reason: prescaled, vetoed, pass-through? last successful step in each chain?• Navigation: which trigger object caused the trigger decision ?• Trigger Event Data: rerun the trigger selection offline with tightened criteria

Encoded trigger decision

Decodedtrigger menu

Persistence

ESDEvent Summary Data

AODAnalysis Object Data

Decoded Trigger Menu

Configuresfor Datataking

Trigger Information(transient)

Transient


Analysis and Trigger Studies

Common: check if event passed a desired triggerTrigger efficiency, luminosity calculations (on TAG DB)

Navigate to trigger object that caused trigger-acceptZee trigger + single electron trigger to study electron trigger efficiency

Rerunning the trigger without reconstruction of trigger objectsRun trigger as during data taking, but switch off trigger feature extraction (FEX) algorithms (see Simon’s talk)Perform selection with tighter requirements (HYPO algorithms)Turn on curves, …

Most trigger analyses possible on AOD data!


Interaction with the TriggerDB – TriggerTool

Integrated consistency

checking

Tree view for trigger menus(and subsets) Table view for plain data

Simple and advanced search capabilities

Shifter ModeChange prescale factors or trigger menu to react on changing detector or beam conditions

User ModeBrowse: trigger menus, detailed information like algorithm parameters

Expert modeUpload new and manipulate existing configurations

Intuitive JAVA based GUI to browse andmanipulate trigger configurations

Search results

Edit pane (e.g. for chains)


Commissioning LVL1 Configuration

ATLAS Central Trigger Processor (CTP) tested for over a 1 year on cosmic data using input from various detector sub-systems

See also Commissioning the ATLAS trigger by J. Boyd (Wed 17:30 Online Computing)

Cosmic-ray RPC impact points, extrapolated to ground level

ATLAS access shafts

Complete muon cosmic ray slice LVL1 + LVL2 in February 2007

Muon trigger LVL1 provides trigger and seed for LVL2 LVL2 algorithm requests detector data to reconstruct muon candidates

Configuration of CTP from the TriggerDB is default

August commissioning week:Writing of LVL1 trigger configuration to conditions database for each run


Commissioning HLT Configuration

Large Scale Tests in December 2006Simple trigger setup, large computing farm: 600 dual core processorsTriggerDB access via the ATLAS DbProxy – service that caches database requests and replies to reduce load on primary database

Two following technical runs in March and May 2007 on simulated and cosmic data

Small computing farm (part of the final ATLAS trigger farm)Test of complex trigger menusExercising the TriggerTool in a shift-like environment

August commissioning week:Tested writing of HLT configuration to conditions database for each run

ATLAS Run Control interface during LST


Trigger Configuration for Simulation Production

AdvantagesConsistency between online trigger and simulation

Configurations created and used during data taking can easily be used in MC production

More flexible propagation of configuration changes to MC production sites

Currently a new MC production cycle requires a new software releaseCan be replaced by software release + configuration key faster turn around if only trigger menus or algorithm parameters need adjustment

System under construction

TriggerDB holds complete trigger configuration can be used to simulate the trigger in the exact same way as it is configured for data taking


Summary

System designed and implemented to provide consistent configuration of all three trigger levels

Access to trigger configuration consistently for data taking, Monte Carlo production, and trigger studies / trigger aware analysis

Fast and flexible response to changing detector or beam conditions during data taking

System provides a trigger history over the ATLAS lifetime, easily accessible by the analyst using the TriggerTool

Commissioning of the system for Level 1 and HLT separately and combined during cosmic test runs and large scale farm tests

Documents

The ATLAS Trigger Configuration System Design and Commissioning A.dos Anjos, P.Bell, D.Berge, J.Haller, S.Head, T.Kohno, S.Li, T.McMahon, M.Nozicka, H.v.d