david n. brown lawrence berkeley national lab representing the babar collaboration the babar mini ...

14
David N. Brown Lawrence Berkeley National Lab Representing the BaBar Collaboration The BaBar Mini BaBar BaBar’s Data Formats Design of the Mini Mini Performance and Status The Mini in BaBar’s New Computing Model

Upload: marylou-tate

Post on 13-Jan-2016

216 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: David N. Brown Lawrence Berkeley National Lab Representing the BaBar Collaboration The BaBar Mini  BaBar  BaBar’s Data Formats  Design of the Mini

David N. Brown

Lawrence Berkeley National Lab

Representing the BaBar Collaboration

The BaBar Mini

BaBar

BaBar’s Data Formats

Design of the Mini

Mini Performance and Status

The Mini in BaBar’s New Computing Model

Page 2: David N. Brown Lawrence Berkeley National Lab Representing the BaBar Collaboration The BaBar Mini  BaBar  BaBar’s Data Formats  Design of the Mini

David N. Brown LBNL BaBar 2 CHEP03 25 March, 2003

BaBar

5-layer Si Vertex tracker40-layer Drift ChamberDIRC Cherenkov CounterCsI Crystal CalorimeterMuon Chambers in Fe

Page 3: David N. Brown Lawrence Berkeley National Lab Representing the BaBar Collaboration The BaBar Mini  BaBar  BaBar’s Data Formats  Design of the Mini

David N. Brown LBNL BaBar 3 CHEP03 25 March, 2003

BaBar Event Data Formats (2001)

1 These formats were not used except by a few experts2 The ESD format was not complete in 2001, supporting only 3 of the 5 Babar subsystems

FormatDesign Size

(Kbytes)Actual Size

(Kbytes)Usage (2001)

Online raw data 25 32 Reconstruction input

Objectivity raw data 25 50 Unused1

Reconstruction output 100 120 Unused1

Event Summary (ESD) 10 7 Unused1 2

Micro 1 3 Physics Analysis

Event Tags 0.1 1 Event Selection

Event Header ? 4 Component navigation

Page 4: David N. Brown Lawrence Berkeley National Lab Representing the BaBar Collaboration The BaBar Mini  BaBar  BaBar’s Data Formats  Design of the Mini

David N. Brown LBNL BaBar 4 CHEP03 25 March, 2003

A Gap in the Formats Raw, Reco, and ESD formats were not useful

Reading any of these required staging many large files Time to read reco is comparable to running reconstruction

A better persistent model was needed: the Mini

Reco

Raw

Reco Track

Kalman Fit

DC Hit Si Hit Si Hit

Cluster

digi

Cluster

digi digi digi

Kalman Fit

Reco Track

Kalman Fit

DC Hit Si Hit Si Hit

Cluster

digi

Cluster

digi digi digi

Kalman FitESD

ESD+

Reco

Transient Persistent

Page 5: David N. Brown Lawrence Berkeley National Lab Representing the BaBar Collaboration The BaBar Mini  BaBar  BaBar’s Data Formats  Design of the Mini

David N. Brown LBNL BaBar 5 CHEP03 25 March, 2003

Mini Design Goals Support detector studies

Provide low-level detector details to support common tasks calibration, alignment, diagnostics, and algorithm development

Provide Reconstruction Object interfaces Support detailed Physics analysis

Provide access to the full reconstruction resultsEG: track fits using Kaon mass for material effect predictions

Allow users to follow calibration and alignment changes Allow detector-level systematic error checks Support a detailed event display Support the standard BaBar analysis interface

Make it easy to access A disk size of 10KBytes/event or less A readback speed comparable to the Micro Allow customized output to fit specific needs

Page 6: David N. Brown Lawrence Berkeley National Lab Representing the BaBar Collaboration The BaBar Mini  BaBar  BaBar’s Data Formats  Design of the Mini

David N. Brown LBNL BaBar 6 CHEP03 25 March, 2003

Mini Design Directly persist high-level reconstruction objects

Tracks, calorimeter clusters, PID results, …

Indirectly persist lower-level reconstruction objects Track hits, calorimeter crystals, …

Store ‘raw’ detector quantities (where possible) Digitization values, electronic channel id, …

Pack data to detector precision Aggressively filter detector noise Avoid overhead in low-level ‘persistent’ classes

Used fixed-size classes Align all data members No virtual functions in low-level classes

Page 7: David N. Brown Lawrence Berkeley National Lab Representing the BaBar Collaboration The BaBar Mini  BaBar  BaBar’s Data Formats  Design of the Mini

David N. Brown LBNL BaBar 7 CHEP03 25 March, 2003

Mini Persistence

Reco Track

Kalman Fit

DC HitSi Hit Si Hit

Cluster

digi

Cluster

digi digi digi

Kalman Fit

Transient

Pack data from low-level classes into compact objects Persist the entire transient tree in one persistent object

References become indices into embedded arrays Every event fully described by 13 persistent objects

Persistent

Page 8: David N. Brown Lawrence Berkeley National Lab Representing the BaBar Collaboration The BaBar Mini  BaBar  BaBar’s Data Formats  Design of the Mini

David N. Brown LBNL BaBar 8 CHEP03 25 March, 2003

Data Packing Digitize floating point values

Eg track fit parameters and covariance matrix Set packing precision at ~1% of detector resolution Use locally flat, globally logarithmic packing algorithm

Packing precision depends on the value being packedSupports histograms without binning artifacts

Bitwise OR small fields into packed data words

Track Impact parameter

Pack into 17 bits

longword alignment

Pack parameter error into remaining 15 bits

Page 9: David N. Brown Lawrence Berkeley National Lab Representing the BaBar Collaboration The BaBar Mini  BaBar  BaBar’s Data Formats  Design of the Mini

David N. Brown LBNL BaBar 9 CHEP03 25 March, 2003

Mini Analysis Interface The BaBar Analysis Framework is entrenched

Huge investment in Physicist code after 3 years of operation

The original design supported multiple data formats But it had evolved to depend on details of the Micro Providing Mini-compatibility was a major effort

Changes in the base classes, new subclasses, …

The BaBar Mini Analysis interface is now working Physics (Micro) objects are built from native Mini objects Fully compatible with existing user code Provides access to most Mini-specific features Performance is comparable to reading the Micro

Page 10: David N. Brown Lawrence Berkeley National Lab Representing the BaBar Collaboration The BaBar Mini  BaBar  BaBar’s Data Formats  Design of the Mini

David N. Brown LBNL BaBar 10 CHEP03 25 March, 2003

Mini Performance Data size (after ootidy + gzip compression)

6.5 Kbytes per generic physics event 10 Kbytes per multi-hadron event (~10 tracks)

Readback speed 20 mSec per generic event (1GHz pentium III Linux)

Readback operation % time

Transient creation + deletion 35

Objectivity data read 1030

Physics object creation 20

Framework overhead 10

Data field unpacking 0.1

Page 11: David N. Brown Lawrence Berkeley National Lab Representing the BaBar Collaboration The BaBar Mini  BaBar  BaBar’s Data Formats  Design of the Mini

David N. Brown LBNL BaBar 11 CHEP03 25 March, 2003

Single Event Display- MiniSingle Event Display- Micro

Page 12: David N. Brown Lawrence Berkeley National Lab Representing the BaBar Collaboration The BaBar Mini  BaBar  BaBar’s Data Formats  Design of the Mini

David N. Brown LBNL BaBar 12 CHEP03 25 March, 2003

Mini Status The Mini was released for production in 2002

All components of BaBar detector representedTrigger, MC truth matching, Particle ID, …

‘2002’ Reprocessing is nearly complete Should finish this week! Will provide the full Mini for BaBar’s full data sample Reco, Raw, and ESD were not written in this processing

A Large savings in cpu, IO, tape, lock traffic, …

Mini data is available at BaBar Tier-A sites The total Mini sample will be ~10 TeraBytes Access is through dynamic staging Small samples can be exported to smaller sites Physicists are starting to use it

Page 13: David N. Brown Lawrence Berkeley National Lab Representing the BaBar Collaboration The BaBar Mini  BaBar  BaBar’s Data Formats  Design of the Mini

David N. Brown LBNL BaBar 13 CHEP03 25 March, 2003

BaBar’s New Computing Model BaBar has recently revised its computing model

Prompted (partly) by the need to integrate the Mini

The Mini will be ported to use RootIO Allows interactive (CINT) access to production output Embedded arrays will be converted to Root columns

Embedded object classes will be directly reused

A Reduced Mini will replace the existing Micro Cluster Mini objects used directly in analysis into New Micro Cluster other objects separately to complete the Mini

The Analysis interface will be re-implemented Optimized for Mini access

Page 14: David N. Brown Lawrence Berkeley National Lab Representing the BaBar Collaboration The BaBar Mini  BaBar  BaBar’s Data Formats  Design of the Mini

David N. Brown LBNL BaBar 14 CHEP03 25 March, 2003

Conclusions BaBar has implemented a new Event format: the Mini

Replaces inefficient Raw, Reco, and ESD formats Provides access to detector detail for average users

The full BaBar data sample will soon be available in Mini format

BaBar is implementing a new Computing Model The Mini will be ported to RootIO The Micro will be replaced with a Reduced Mini The Analysis interface will be re-implemented The new model will be deployed in late 2003

We are close to achieving our original goal of a flexible, unified, efficient event data format