1. ALICE Computing Data Challenge VI
Fons Rademakers
Good afternoon.
We will present you today of the ALICE Data Challenge.
This presentation is divided in 5 parts:
I‘ll present you
- the data throughput needed in the ALICE experiment
- the motivations and the history of the data challenge
Roberto Divia will present you the part of the DAQ part of the Data Challenge
Fons Rademakers will present the Input/Ouptut done with the ROOT package
and Bernd Panzer will cover the setup done in the computing center for this test
At the end I will draw conclusions and present the future plans

2. Computing Data Challenge Motivations
Integrate prototypes of components of the DAQ/HLT/Offline system into a complete system covering the whole chain
Measure the performances with existing software prototypes
Step-wise performance and reliability progress towards the final system
Establish a common ALICE test-bed dataflow chain:
Using common LHC test-bed to minimize investment
Test of new architectural ideas and new components
Start the integration now and verify the scalability
Confront the implementation with realistic test conditions
Started in 1999 and to be repeated annually till LHC startup
Motivations are the Data Challenge are the following: -
Establish a common ALICE test-bed
First Data Challenge was started in March 1999
Sept 2004
ALICE Data Challenge VI

3. ALICE Data Challenge VI
Data Challenge Plans
MBytes/s
Sept 2004
ALICE Data Challenge VI

4. Results of Previous CDC’s
Five Data Challenges performed so far:
March 1999
CDC I: 30 MB/s peak, 7 TB to HPSS
March/April 2000
CDC II: 100 MB/s peak, 23 TB to HPSS/CASTOR
February/March 2001
CDC III: 120 MB/s peak, 110 TB to CASTOR
October/November 2002
CDC IV: 300 MB/s peak, 200 MB/s to CASTOR
Dummy writer, no objectifier
December 2004
CDC V: 300 MB/s to CASTOR, but not sustained
Sept 2004
ALICE Data Challenge VI

5. CDC VI Performance Goals
1.5 GB/s through event building network
450 MB/s sustained to tape
300 TB to tape (one week at 450MB/s)
Maximum 150 TB of tape storage
Sept 2004
ALICE Data Challenge VI

6. CDC VI Technology Goals
Testing and validation of DATE modifications
see Pierre’s talk
Test on new CASTOR
see Olof’s talk
Online physics monitoring using HLT algorithms
see Cvetan’s talk
Near real-time off-site data distribution via gLite
Stress test new Enterasys 10 GB switches
Stress test SLC3 and Linux 2.6 kernel
Stress test xrootd
Sept 2004
ALICE Data Challenge VI

7. CDC Schematics DAQ AliMDC CASTOR gLite AliRoot ROOT I/O Raw Data
GEANT3 GEANT4 FLUKA
AliRoot
Raw Data
Simulated Data
DAQ
gLite File Catalogue
Performance Monitoring
ROOT I/O
CERN TIER 0 TIER 1
AliMDC
CASTOR
Regional
TIER 1 TIER 2
gLite
Sept 2004
ALICE Data Challenge VI

8. ALICE Data Challenge VI
High Throughput Prototype (Openlab + LCG prototype)
(specific layout, September 2004)
4 * GE connections
to the backbone
10GE WAN connection
lxsharexxxd
ENTERASYS N7
10 GE Switches
20 Disk Server
(dual P4, IDE disks,
~ 1TB disk space each)
10 GE per node
10GE
10 GE per node
oplapro0xx
tbed00xx
IA32 CPU Server
(dual 2.4 GHz P4, 1 GB mem.)
1 GE per node
2 * 50 Itanium Server
(dual 1.3/1.5 GHz Itanium2,
2 GB mem)
lxs50xx
20 TB , IBM StorageTank
Sept 2004
ALICE Data Challenge VI

9. ALICE Data Challenge VI
High Throughput Prototype (Openlab + LCG prototype)
(specific layout, October 2004)
10GE WAN connection
4 * GE connections
to the backbone
4 *ENTERASYS N7
10 GE Switches
2 * Enterasys X-Series
lxsharexxxd
36 Disk Server
(dual P4, IDE disks,
~ 1TB disk space each)
oplapro0xx
10 GE per node
10 GE per node
tbed00xx
2 * 100 IA32 CPU Server
(dual 2.4 GHz P4, 1 GB mem.)
10GE
1 GE per node
2 * 50 Itanium Server
(dual 1.3/1.5 GHz Itanium2,
2 GB mem)
lxs50xx
20 TB , IBM StorageTank
12 Tape Server
STK 9940B
Sept 2004
ALICE Data Challenge VI

10. September October November December
CPU
Nodes
IA32
IA64 25 30
120 60
200 60
ATA SERVER (25TB)
SATA SERVER (25TB)
Disk
Server
FC SATA (30 TB)
iSCSI ST (28TB)
Tape
Drives
STK 9940B
IBM
LTO2 ?
ALICE MDC Milestone
September October November December

11. CDC DATA Flow LDC (Local Data Concentrator) Gigabit/10 GB
Ethernet Switch
GDC
(Global Data Collector)
tbedXXXX
tbedXXXX
GE Switch
readout
recorder
gdcServer
eventBuilder
alimdc
Half-size form factor 6U
+ DUAL port RAM
CASTOR
Sept 2004
ALICE Data Challenge VI

12. AliMDC - the Objectifier
ROOT Objectifier reads raw data via shared memory from the DATE GDC
Objectifier has two output streams:
Raw event database (to CASTOR)
File catalog to (to gLite)
25-30 MB/s raw database output per 2 GHz PC via rootd to CASTOR
Hook for HLT and physics monitoring
HLT algorithm will add about s/event
Try to multithread algorithm
Adds ESD data to RAW file
Add hook to simulate level 1 input
Add some level 1 info to event header object
Sept 2004
ALICE Data Challenge VI

13. ALICE Data Challenge VI
Raw Data
All detectors except RICH, ZDC and FMD can generate raw data
Used to produce raw data for input to DATE LDC’s
Sept 2004
ALICE Data Challenge VI

14. Regional Center Participation
Certain files will be copied to a disk cache
Regional centers will be able to access these files online via gLite
Offline, regional centers can copy/access files, via gLite, from CASTOR
Candidate regional centers?
Sept 2004
ALICE Data Challenge VI

15. ALICE Data Challenge VI
To Be Done
AliMDC
Finalize and test TCastorFile using new CASTOR API
Interface xrootd to new CASTOR API
Add gLite interface
HLT
See Cvetan’s talk
gLite
To be deployed
Regional Centers
Verify infrastructure
Sept 2004
ALICE Data Challenge VI

16. ALICE Data Challenge VI
Timeline
AliMDC ready: early Oct 2004
HLT: Oct 2004
gLite ready: Oct 2004
Regional centers ready: Nov 2004
Integration run with online: end Oct 2004
CDC VI: second half Nov till first half Dec 2004
Sept 2004
ALICE Data Challenge VI