october 20 th, 2000lyon - daq2000hp beck atlas trigger & data acquisition requirements and...

October 20th, 2000 Lyon - DAQ2000 HP Beck

ATLASATLASTrigger & Data AcquisitionTrigger & Data Acquisition

Requirements and ConceptsRequirements and Concepts

ATLASATLASTrigger & Data AcquisitionTrigger & Data Acquisition

Requirements and ConceptsRequirements and Concepts Hanspeter Beck

LHEP - Bernfor the ATLAS T/DAQ Group

DAQ 2000Workshop on Network-Based Data Acquisition

and Event-Building

at the Nuclear Science Symposium

andMedical Imaging Conference

22


OverviewOverviewOverviewOverview

• LHC characteristicsLHC characteristics• The ATLAS experimentThe ATLAS experiment• Requirements for Trigger DAQRequirements for Trigger DAQ• T/DAQ architectureT/DAQ architecture• Networking for DAQ tasksNetworking for DAQ tasks• Project StatusProject Status

• LHC characteristicsLHC characteristics• The ATLAS experimentThe ATLAS experiment• Requirements for Trigger DAQRequirements for Trigger DAQ• T/DAQ architectureT/DAQ architecture• Networking for DAQ tasksNetworking for DAQ tasks• Project StatusProject Status

33


The Large Hadron Collider at CERNThe Large Hadron Collider at CERNThe Large Hadron Collider at CERNThe Large Hadron Collider at CERN

Startup of LHC: 2005Startup of LHC: 2005

44


LHCLHC CharacteristicsCharacteristicsLHCLHC CharacteristicsCharacteristics

Interaction rate (ATLAS, CMS) Interaction rate (ATLAS, CMS) 10 1099 HzHzInteraction rate (ATLAS, CMS) Interaction rate (ATLAS, CMS) 10 1099 HzHz

• LHC circumference 26.7 km ~100 m underground

• Center of mass energy 14 TeV (i.e. 7 TeV per beam)

• Protons per bunch 0.17 ·1011 (1.67 ·1011 for high Luminosity)

• Number of bunches 3564 (of which 2835 are filled)

• Size of a bunch radius σx = σy = 16 mlength = 56 mm

• Spacing between bunches 7.48 m 24.95 ns 40 MHz

• Interactions per bunch 23 minimum-bias events (high Luminosity)

• Experiments ALICE, ATLAS, CMS, LHCb

55


ATLAS LuminosityATLAS LuminosityATLAS LuminosityATLAS Luminosity

Peak 1033 cm-2 s-1 2005-2008 (“low luminosity”)

Peak 1034 cm-2 s-1 2008 (“high luminosity”)

dt 10 fb-1 per year at low luminosity

dt 100 fb-1 per year at high luminosity

Bunch crossing: 25 ns ~ 23 minimum-bias /crossing at high luminosity (pile-up)

Detector speed Radiation hardness Trigger selection and data acquisition

66


The ATLAS ExperimentThe ATLAS ExperimentThe ATLAS ExperimentThe ATLAS Experiment

77


ATLAS Main ComponentsATLAS Main ComponentsATLAS Main ComponentsATLAS Main Components

Magnet(s)

Air-core toroidssolenoid in inner cavityNo field in calorimeters4 magnets

TRACKER

Si pixel + stripsTRT particle IDB = 2 T/pT ~ 5x10-4 pT 0.01

9 314 304 channels424 576 “

EM CALOLiquid Argon/E ~ 10%/E uniformlongitudinal segmentation

173 952 “

HAD CALOFe-scint. Tiles (+LAr ~ 10 )/E ~ 50%/E 0.03

25 714 “

MUON Air /pT ~ 10% at 1 TeVstandalone

1 221 096 “

1600 ReadOut Links1600 ReadOut Links 2.2 Mbyte Event Size2.2 Mbyte Event Size1600 ReadOut Links1600 ReadOut Links 2.2 Mbyte Event Size2.2 Mbyte Event Size

88


Physics MotivationsPhysics MotivationsPhysics MotivationsPhysics Motivations

• Origin of masses and EW symmetry breaking– Look for a Standard Model Higgs

– Final word about SM Higgs mechanism

• Physics beyond the Standard Model– SUSY : explore up to masses of ~ 3 TeV

– Final word about low-energy SUSY

– Other scenarios: leptoquarks, technicolor, ...

– Additional /q/w/z, etc. Up to m ~ 5 TeV

• Precision measurements – W, TGC, top

– QCD

– B-physics and CP violation

99


ATLAS Three Trigger LevelsATLAS Three Trigger LevelsATLAS Three Trigger LevelsATLAS Three Trigger LevelsCalorimeter + Muon coarse trigger data

Region of Interestfull granularityFull event reconstructionaccess to latest calibration

and alignment tables

10-4

10-2

100

102

104

106

108

10-8 10-6 10-4 10-2 10-0

Rate [Hz]

25 ns s ms sec

LVL1 40 MHz

LVL2 75 kHz (100 kHz)

Event Filter O(1) kHz

Storage O(100) Hz

Jetsb /K

W, Z

t

H

< 2.5 s

<10> msseconds

Processing Time [s]

1010


FE channels

ReadOut Buffers

FE Links

ReadOut Links

O(1) GB/s

40 MHz 1 GHz

O(1) kHz

75 kHz (100 kHz)

EF Farm

Event Building

SFI

O(100) Hz

Level-2 Trigger System

Full Event Building

ROD

<2.5 s

RoI pointers

RoI Data

L2 Acc/Rej

~10 ms

Event Filter

Farm ~sec

O(100) GB/s

ATLAS DAQ and TriggerATLAS DAQ and TriggerATLAS DAQ and TriggerATLAS DAQ and Trigger

SFO O(100) MB/s

ReadOut Drivers

O(100) GB/sRequest/Receive

Mass Storage

ROS

RO

B

RO

B

RO

B

ReadOut System

1111


Read Out

BuffersTR

G

EB

IF

Loc

alC

ontr

olle

r

LVL2

Read Out

BuffersTR

G

EB

IF

ROI builder

EVENT BUILDER

Event Filter

Mass Storage

. . . .

SFI

SFO

. . . SFI

SFO

Online s/w

RunControlConfigureMonitoring

LVL1 Detector 1 Detector 2 Detector n

Loc

alC

ontr

olle

rL

ocal

Con

trol

ler

Loc

alC

ontr

olle

r

Loc

alC

ontr

olle

r

Loc

alC

ontr

olle

r

LVL2

ROI builder

TR

G

TR

G

. . . .

LVL2

ROI builder

. . . .

Detector 1 Detector 2 Detector n

Read Out

Buffers

Read Out

Buffers

LVL1

RoI

RoI

LVL2

. . . .. . . .EB

IF

EB

IF

EVENT BUILDER

TR

G Read OutBuffers T

RG Read Out

Buffers

EVENT BUILDER

SFI

Event Filter

SFI

SFO

Event Filter Mass Storage

SFO

1212


LVL2 DataFlowLVL2 DataFlowLVL2 DataFlowLVL2 DataFlowTotal bandwidth in LVL2 network ~ 5 Gbyte/s

Max output bandwidth to LVL2 per ROB ~ 9 Mbyte/s

Traffic @ ROB [request-response/sec] ~ 11 kHz

Fragment size per ROB ~ 1 kbyte

Typical number of ROBs per data request ~ 4

Average number of ROBs required to supplydata per RoI

10 – 35

Typical number of RoIs per event 1 – 2

Number of ROBs ~ 1600

Max RoI Builder / Supervisor rate 75 (100) kHz

Worst case assumption according B-Physics selection at LVL2; Worst case assumption according B-Physics selection at LVL2; intended for low Luminosity only.intended for low Luminosity only.Worst case assumption according B-Physics selection at LVL2; Worst case assumption according B-Physics selection at LVL2; intended for low Luminosity only.intended for low Luminosity only.

1313


LVL2 DataFlowLVL2 DataFlowLVL2 DataFlowLVL2 DataFlow

LVL2DataFlow

LVL1 Trigger ReadOut System

EventBuilding

Online SoftwareLVL2 Selection

RoI & LVL1 data

Event dataEvent data request

LVL2 decisions

RunControlconfiguremonitoring

Event data requests

LVL2 decisions

RoI & LVL1 data

Requested event data

LVL2 accept

LVL2 DataFlow Context Diagram

1414


LVL2 DataFlowLVL2 DataFlowLVL2 DataFlowLVL2 DataFlow

For prototype implementationsFor prototype implementationstestbed measurementstestbed measurementsand modeling activities for the LVL2 and modeling activities for the LVL2 dataflow:dataflow:

See the talks of See the talks of Denis CalvetDenis Calvet and and Micheal Le Vine Micheal Le Vine later this afternoon.later this afternoon.

For prototype implementationsFor prototype implementationstestbed measurementstestbed measurementsand modeling activities for the LVL2 and modeling activities for the LVL2 dataflow:dataflow:

See the talks of See the talks of Denis CalvetDenis Calvet and and Micheal Le Vine Micheal Le Vine later this afternoon.later this afternoon.

1515


EventBuilding DataFlowEventBuilding DataFlowEventBuilding DataFlowEventBuilding DataFlow

Total bandwidth in EB network 1- 8 Gbyte/s

Max output bandwidth to EB per ROS 1- 75 Mbyte/s

Fragment size per ROS 1- 15 kbyte

Fragment size per ROB ~ 1 kbyte

Number of ROBs per ROS 1- 15 ROBs

Number of ROSs 100- 1500

Number of SFIs 100- 200

LVL2 accept rate 1- 5 kHz

1616


EventBuilding DataFlowEventBuilding DataFlowEventBuilding DataFlowEventBuilding DataFlow

EventBuilding

Trigger ROS

Online SoftwareSFI

Accept Event Control

Event fragments

RunControlconfiguremonitoring

EventBuilding Context diagram

Event fragments

Control

1717


Event Builder ModelEvent Builder ModelEvent Builder ModelEvent Builder Model

DFM

ROS

SFI

LocalController

Trigger

Data transfer

End of Event, Busy/NonBusy

Destination assignment

Run control

DFM DataFlow Manager

EoE, B/B

1818


DataFlow - Two-layer ApproachDataFlow - Two-layer Approach Split Functionality and Technology

DataFlow - Two-layer ApproachDataFlow - Two-layer Approach Split Functionality and Technology

Upper layer:

OS and technology independencefunctionality: LVL2 DataFlow + Event Building

data + control messages

Lower layer:

technology dependentfunctionality: data transfer

Decoupling of upper and lower layer:

with technology independent API anddifferent technology implementations

ATM /AAL5

TCP /IP

...

Message Passing

Appl Appl Appl

Baseline candidate technology is Fast+Gigabit Ethernet using various protocols on it, from raw frames (MESH) up to TCP/IP.

(Other technologies have been studied and could still be resurrected, e.g. ATM, FC)

1919


EventBuilder TestbedsEventBuilder TestbedsEventBuilder TestbedsEventBuilder Testbeds

Fast Ethernet &ATMGigabit Ethernet

2020


EventBuilder Event RatesEventBuilder Event Rates EventBuilder Event RatesEventBuilder Event Rates

Simulation(Ptolemy model)

MeasuredPerformance(Gigabit Ethernet TCP/IP)

0

5

10

15

20

25

30

0 5 10 15 20 25 30 35

Fragment size [kByte]

Eve

nt

rate

[kH

z]

4x4 EB on ATM 155 Mbit/s - data

4x4 EB on ATM 155 Mbit/s - simulation

6x6 EB on Gigabit Ethernet - data

Average event size = 2.2 MbyteATM 155 Mbit/s

0

1

2

3

4

100 200 400 600

Number of Sources = Number of Destinations

Eve

nt

rate

[kH

z]

2121


Status of the ProjectStatus of the ProjectStatus of the ProjectStatus of the Project

DAQ/EF -1 (1996-2000)DAQ/EF -1 (1996-2000)

Prototype implementinga full slice of DAQ system(excluding LVL2 trigger).Emphasis on system aspects,i.e. full functionality incl. configuration and monitoring fromROB, EB, SFI, EF to SFO andstorage.

Was successfully used for ATLAS testbeam this year.

Pilot Project (1998-Pilot Project (1998-2000)2000)

Based on previous demonstrator programs, it aimed on proving the principle of function of LVL2. Emphasis on triggeraspects i.e. implications ofRoI concept and exploiting the full performance potential ofnetworks i.e. usage of raw frames, drivers...

In the past few years DAQ aspects have been studied separately from LVL2 (trigger and dataflow). Two groups were since working in parallel, exploiting feasibility of their respective system aspects:

2222


Status of the ProjectStatus of the ProjectStatus of the ProjectStatus of the Project

The outcome of both projects enabled us to define the architecture of the final TDAQ system (March 2000).

HLT, DAQ and DCS Technical Proposal (LHCC/2000 - 17)

Currently, the fusion of the DAQ/EF -1 prototype and the Pilot project into an integrated prototype is in planning. A testbed running the full TDAQ architecture is expected to be exploited by summer next year.

The Technical Design Report (TDR) will be based on the assessments of this integrated prototype.

Only then, the final TDAQ system will be built according to the needs of the assembly of the ATLAS detector.

2323


SummarySummarySummarySummary

• The experiments at LHC will allow us to shed light on the particle mass generation mechanism; to do precision measurements on many parameters of the Standard Model; and to peek into new physics domains beyond the current Standard Model of particle physics.

• The high luminosity at LHC and the size of the ATLAS detector require the development of a sophisticated data acquisition system, with online event selection.

• The efficient use of high performance (low latency, high throughput) networks will play a key role in the success of LHC.

october 20 th, 2000lyon - daq2000hp beck atlas trigger & data acquisition requirements and...

Documents

atlas experiment slide

atlas luminosity peak

atlas tdaq group daq

data acquisition slide

trigger daqrequirements

lhcb slide

trigger selection

trigger system