BaBar RAL Staff Talks. 4th April 2005 RAL BaBar Staff Talk 2 Outline The Group Fergus Wilson – Group Leader since Oct 2004 Tim Adye Emmanuel Olaiya.

BaBar RAL Staff Talks

4th April 2005

RAL BaBar Staff Talk 2

Outline The Group

Fergus Wilson – Group Leader since Oct 2004 Tim Adye Emmanuel Olaiya (at SLAC) Chris Brew (50% Computing Group) Boda Franek (50% LHCb) Adil Hasan (briefly!)

The Speakers: Fergus – RAL, BaBar Physics and the Future. Tim – BaBar Computing as a testbed for the LHC. Chris – BaBar and the GRID. Adil – New Computing Technologies (SRB).

Non-Speakers Boda - Run Control. Gian - Retired but still working on physics. Manny - Our contact person at SLAC.

4th April 2005


In the beginning…Why does the Universe look like this… …and not like this?

Perhaps matter and anti-matter behave differently…

4th April 2005


Stanford Linear Accelerator CenterSea (Pacific) Sand Surf

San Francisco

Snow

San Jose

and some physics…

Sun

San Andreas Fault

10 Countries 77 Institutions 593 Physicists

5

The BaBar Detector at SLAC

Linear Accelerator

2.2km ring BaBar

26th May 1999: first events recorded by BaBar

4th April 2005


The Accelerator

250 million collisions per second from which we store on disk about 5 per second.

Currently have 230 million B pairs (about 18 times previous experiments)

World’s largest computer database (ignoring the NSA…)

BBSee )4(

4th April 2005


What RAL is involved in Supporting the Experiment

See Tim’s talk Developing the GRID

See Adil’s Talk Developing GRID production for BaBar

See Chris’s Talk Leading the First Level Trigger hardware upgrade

project Completed in 2004 and now running (Nicolo de Groot)

Exploiting the physics Next page…

4th April 2005


We have concentrated on charmless decays, direct CP, Branching Fraction measurements, interference, extraction of weak and strong phases.

Analyses inherited from Bristol: Dalitz Analysis (RAL/Bristol/Birmingham/Warwick)

B+ → π+π+π-

B+ → K+π+π-

B+ → K+K+K-

B+ → a1+π0, B+ → a1

0π+ (Dan Walker, student) B+ → K*0K- , B+ → K+K-π- Dalitz (Jim Burke, student) Lepton Flavour Violation: e+e- → μ+τ- (Marc Kelly)

New and upcoming analyses: B+(0) → K*+(0) h- h+ (inclusive/exclusive decays) (Manny + Fergus) Lepton Flavour Violation: e+e- → μ+τ- (Gian Gopal + Fergus) B+ → K*+π0 , B+ → K0

S ρ+ (Fergus) Factorization fits (Noel Cottingham + Fergus)

If we can pull this off, we will have ~9 papers with major RAL input.

2005 RAL Physics

I should have written the first paper last week…

4th April 2005


The Future BaBar should run until 2008. Quadruple our dataset. 1/3rd of the collaboration (200 people?) will do their analysis at RAL. RAL will continue to be central to BaBar’s exploitation for many years to

come.

BaBar Tier A at BaBar Tier A at RALRAL

Tim Adye

PPD Staff Talks4th April 2005

4th April 2005 RAL BaBar Staff Talk 11

RAL Tier A

• RAL is a BaBar “Tier-A” site• “Tier-1” in Grid-speak• This means that we provide a good fraction of the

Collaboration’s computing resources• in exchange for a rebate from the UK’s contribution

• Until last year, it was just SLAC, Lyon, and us.• Now also Karlsruhe and Bologna


Our Rôle

• RAL has concentrated on physics analysis computing• Other sites took care of generic production

• Event Reconstruction, Monte Carlo generation, etc

• Has meant lots of physicist users (277 in total)


• Tim, Manny, and Chris support BaBar physicists using RAL

• Rely on the excellent service from the Tier-1/A support staff, in particular

• Martin Bly and Steve Traylen (PPD)• Nick White and Andrew Sansum (eScience)


What we have to do

• Keep software up to date• New BaBar releases every few weeks (1-2 GB each)• Maintain BaBar-specific services

• Objectivity, xrootd, etc

• Customise the user environment• Make it as SLAC-like as possible

• Import and manage BaBar data• events and detector conditions• Currently have 70 TBytes, but we’ll run out soon!

• Register new BaBar users• Help our user community with problems

• most are specific to BaBar, so need expertese beyond just system support


BaBar Batch CPU Use at RAL

0

50,000

100,000

150,000

200,000

250,000

Week Beginning

BaB

ar C

PU

Ho

urs

per

Wee

k(N

orm

alis

ed to

P45

0)

SPUK UsersNon-UK Users


BaBar Batch Users at RAL(running at least one non-trivial job each week)

0

5

10

15

20

25

30

35

40

45

50

Week Beginning

BaB

ar U

sers

per

Wee

k

UK UsersNon-UK Users

A total of 284 new BaBar users registered since December 2001


The BaBar Computing Model

• Until last year, BaBar had two event data formats1. Objectivity

• Object-oriented database. Commercial software.• Original design. Used for simulation and reconstruction.• Cumbersome for physics analysis and data distribution

2. KanGA• “Kinder and Gentler Analysis”. Simple ROOT files.• Easier for users, more compact, and simpler to install

and import.

• RAL was the only place that held all the KanGA dataset• This made us very popular!• I developed the import and management tools used

here and at all other (Tier C) sites.


New Computing Model

• This hybrid couldn’t last• Using both Objectivity + KanGA was a waste of resources,

expertese, and time• and Objectivity was a nightmare (if not for us at RAL )

• Last year BaBar developed a new computing model• Dropped Objectivity and improved KanGA• I did the data distribution and user bookkeeping tools

• completely redesigned bookkeeping (“meta-data”) system• this took over the “event navigation” functionality previously

done (very badly) by Objectivity

• Physics analyses now assigned to different Tier As• according to the type of data (“skims”) they need• We are responsible for “Charmless” and “Tau/QED”

• These are predominently UK topics


My other interests (a)

1. “SqlAbstractor”• Came out of my work on the BaBar bookkeeping• Provides an object-oriented interface to a group of

database tables• Perl and command-line interfaces• user does not need to know about table

structure/schema• developer can provide “synthetic” data items or

override database selections/return values• A public release is nearly ready (CPAN)


My other interests (b)

2. Created new Masterclass exercise• replaces 8 year-old OPAL exercise• students view BaBar events to measure B-lifetime• 500 students used it for the first time this year

• Measured τB = 2.6 ps. PDG has 1.6 ps.• Not bad for an hour’s analysis!

3. Starting to look at making a generic “unfolding”(AKA “deconvolution” or “unsmearing”) package for ROOT• Currently there are several techniques

• each implemented in it’s own stand-alone program

• I want to allow people to access them from a single interface• Can compare different methods without starting from scratch• Analagous concept to using MINUIT or RooFit for fitting


My other interests (c)

5. PPD web• with Gareth Smith and Janet Haswell

6. RAL HEP Summer School tutor• though I feel more like a student than a tutor!

7. Science fiction• books, films, criticism

• I even collaborated on a book about it• (pity about the naff cover the publisher gave us)

• … but I guess this is supposed to be aboutmy professional interests.

4th April 2005


Boda Franek

4th April 2005


SMI

Object Oriented Framework for designing and implementingDistributed Control Systems

•Upgraded from SMI to SMI++ in 1996-99 for use in BaBar to design BaBar Run Control ( B.Franek and C.Gaspar)

•Currently also being used by•ATLAS and CMS for the design of their Detector Control System•ALICE and LHCb for the design of the complete Experiment Control System

•Maintained and further developed by B.Franek

4th April 2005


Run Control ORC

Designed and implemented using SMI++ toolkit 1996-1999 by B. Franek

Since 1999 maintained and further developed

Online Run Control

4th April 2005


Chris Brew

4th April 2005


BaBar SPGrid Last round of BaBar MC Production generated

2.7B Events at about 69M Ev/Week Distributed production over 19 sites with person

at each site responsible for maintaining software and running jobs

Next round needs to produce 1.5 – 2 times as much data in a shorter time

Trying to use grid tools to use more resources with less effort

4th April 2005


Challange Idea is to make LCG look just like another farm to

the BaBar Production system so one manager on one machine can submit jobs to many sites

But BaBar MC Jobs have needs: Software Conditions DB Background Input

We need to be able to make sure these are available to the Job

4th April 2005


Status BaBar MC software is now packaged for pre-

installation using LCG tools Interface of LCG job submission and data

recovery into MC production tools is working Simple monitoring of jobs is working Initial shake down production of ~4 Million

events on the RAL Tier 1 and RALPP Tier 2 is running. Peak rate about 500,000 Ev/day

4th April 2005


Deploying Xrootd at RAL BaBar will have 75TB of data at RAL comprising

150k files on 64 disks Real headache keeping track of them Xrootd system developed at SLAC as part of

project to replace Objectivity Two type of servers

Load Balancers Data Servers

Client contacts the Load Balancer which takes care of finding the file they want and redirects them to a data server that has it

4th April 2005


Data Serverxroot daemonolb daemon

Load Balancerolb daemonxrootd daemon

Client

4th April 2005


Xrootd at RAL 2 Now deployed at RAL with 46 TB of data on 21

servers Adding resiliency by interfacing to the Atlas

DataStore. All data is being copied to tape If a server goes down and a client requests a file that

was on that server the client will automatically be redirected to another server that will copy the file from that DataStore

Even works if a server goes down while a client is reading a file

4th April 2005


4th April 2005


Adil Hasan

BaBar and SRB

Adil Hasan (RAL) - project coordinator

Wilko Kroeger (SLAC) and Jean-Yves Nief (ccin2p3)

4th April 2005


What is SRB? The Storage Resource Broker developed by San Diego Super Computing

Center is a Client/Server architecture. Provides uniform interface to different network-connected resources.

So can access tape resource and disk resource in same way. Provides a metadata catalogue.

Allows logical names to be associated to physical files (concept of collections, datasets, etc).

Allows fine-grained access control. Has the concept of Federation to allow multiple SRBs to act as part of the same

Virtual Organization.

4th April 2005


What is SRB?

4th April 2005


SRB in BaBar Part of the BaBarGrid effort. Used for Tier 1 (or A) data distribution from SLAC to ccin2p3,

currently more than 95TB data copied (over 170K files). For new run (starting now) have setup and federate SRB MCAT at

SLAC, ccin2p3 and will setup one at RAL. Will allow SLAC, ccin2p3, RAL to copy data from each other easily

(don’t need to know site storage intricacies). Currently working on incorporating client-side applications into user

data-distribution tools (Tim).

4th April 2005


Fermila

b (US)

CERN

SLAC (US)

IN2P3 (F

R)

1 T

erab

yte/

day

SLAC (US)

INFN P

adva (I

T)

Fermila

b (US)

U. C

hicago (U

S)

CEBAF (US)

IN2P3 (F

R)

INFN P

adva (I

T) S

LAC (US)

U. Toro

nto (CA)

Ferm

ilab (U

S)

Helmholtz

-Karls

ruhe (D

E) S

LAC (US)

DOE Lab D

OE Lab

DOE Lab D

OE Lab

SLAC (US)

JANET (U

K)

Fermila

b (US)

JANET (U

K)

Argonne (U

S) Leve

l3 (US)

Argonne

SURFnet (

NL)

IN2P3 (F

R) S

LAC (US)

Fermila

b (US)

INFN P

adva (I

T)

Example of network utilization on ESNET (US): 1 server.

BaBar RAL Staff Talks. 4th April 2005 RAL BaBar Staff Talk 2 Outline The Group Fergus Wilson – Group Leader since Oct 2004 Tim Adye Emmanuel Olaiya.

Documents