CERNs openlab Project

1
CERNs openlabProject

• Sverre Jarp
• IT Division
• CERN
• 21 November 2002

2
Our ties to IA-64 (IPF)

• A long history already.
• Nov. 1992 Visit to HP Labs (Bill Worley)
• We shall soon launch PA-Wide Word!
• 1994-6 CERN becomes one of the few external
  definition partners for IA-64
• Now a joint effort between HP and Intel
• 1997-9 Creation of a vector math library for
  IA-64
• Full prototype to demonstrate the precision,
  versatility, and unbeatable speed of execution
• 2000-1 Port of Linux onto IA-64
• Trillian project glibc
• Real applications
• Demonstrated already at Intels Exchange
  exhibition on Oct. 2000

3
The openlab advantage

• openlab will be able to build on the following
  strong points
• CERN/ITs technical talent
• CERN existing computing environment
• The size and complexity of the LHC computing
  needs
• CERN strong role in the development of GRID
  middleware
• CERNs ability to embrace emerging technologies

4
Back to openlab

• Industrial Collaboration
• Enterasys, HP, and Intel are our partners
• Technology aimed at the LHC era
• Network switch at 10 Gigabits
• Connect at both 1 Gbit and 10 Gbits
• Rack-mounted HP servers
• Itanium processors
• Storage subsystem may be coming
• from a 4th partner
• Cluster evolution
• 2002 Cluster of 32 systems (64 processors)
• 2003 64 systems (Madison processors)
• 2004 64 systems (Montecito processors)

5
The compute nodes

• HP rx2600
• Rack-mounted (2U) systems
• Two Itanium-2 processors
• 900 or 1000 MHz
• Field upgradable to next generation
• 4 GB memory (max 12 GB)
• 3 hot pluggable SCSI discs (36 or 73 GB)
• On-board 100 and 1000 Mbit Ethernet
• 4 full-size 133 MHz/64-bit PCI-X slots
• Built-in management processor
• Accessible via serial port or Ethernet interface

6
openlab SW strategy

• Exploit existing CERN infrastructure
• Which is based on
• RedHat Linux, GNU compilers
• OpenAFS
• SUE (Standard Unix Env.) systems maintenance
  tools
• Native 64-bit port
• Key LHC applications
• CLHEP, GEANT4, ROOT, etc.
• Important subsystems
• Castor, Oracle, MySQL, LSF, etc.
• Intel compiler where it is sensible
• Performance
• 32-bit emulation mode
• Wherever it makes sense
• Low usage, no special performance need
• Non-strategic areas

7
openlab - phase 1

• Integrate the openCluster
• 32 nodes development nodes
• Rack-mounted DP Itanium-2 systems
• RedHat 7.3 (AW2.1 beta) kernel at 2.4.19
• OpenAFS 1.2.7, LSF 4
• GNU, Intel Compilers ( ORC?)
• Database software (MySQL, Oracle?)
• CERN middleware Castor data mgmt
• GRID middleware Globus, Condor, etc.
• CERN Applications
• Porting, Benchmarking, Performance improvements
• CLHEP, GEANT4, ROOT, CERNLIB
• Cluster benchmarks
• 1 ? 10 Gigabit interfaces

Estimated time scale 6 months Prerequisite 1
system administrator
Also Prepare porting strategy for phase 2
8
openlab - phase 2

• European Data Grid
• Integrate OpenCluster alongside EDG testbed
• Porting, Verification
• Relevant software packages
• Large number of RPMs
• Document prerequisites
• Understand dependency chain
• Decide when to use 32-bit emulation mode
• Interoperability with WP6
• Integration into existing authentication scheme
• Interoperability with other partners
• GRID benchmarks (As available)

Also Prepare porting strategy for phase 3
Estimated time scale 9 months(May be subject to
change!) Prerequisite 1 system programmer
9
openlab - phase 3

• LHC Computing Grid
• Need to understand
• Software architectural choices
• To be made between now and mid-2003
• Need new integration process of selected software
• Time scales
• Disadvantage
• Possible porting of new packages
• Advantage
• Aligned with key choices for LHC deployment

Impossible at this stage to give firm estimates
for timescale and required manpower
10
openlab time line
Order/Install 32 nodes
Systems experts in place Start phase 1
Complete phase 1
Start phase 2
Order/Install G-2 upgrades 32 more nodes
Complete phase 2
Start phase 3
openClusterintegration
Order/Install G-3 upgrades
EDG inter-operability
LCG interoperability
End-02
End-03
End-04
End-05
11
Gelato contributions

• CERN is planning to contribute
• ROOT (C data analysis framework)
• Already on Gelatos Web page thanks to the
  authors
• Castor (data management subsystem)
• GPL Already installed on our cluster
• CLHEP (class library for HEP)
• GPL Already installed/tested
• GEANT4 (C Detector simulation toolkit)
• We need to understand the exact license
  implications
• CERNLIB (all of CERNs FORTRAN software)
• GPL Being ported/installed on Itanium
• Zebra memory banks are I4
• Datagrid software
• GPL-like license

12
IA-64 wishlist

• For IA-64 (IPF) to establish itself solidly in
  the market-place
• Better compiler technology
• Offering better system performance
• Wider range of systems and processors
• For instance Really low-cost entry models, low
  power systems
• More vendors
• State-of-the-art process technology

Similar commoditization as for IA-32
13
The potential of openlab

• Leverage CERNs IT strengths
• Integrate perfectly into our environment
• OS, Compilers, Middleware, Applications
• Integrate alongside EDG testbed
• Integrate into LCG deployment strategy
• Show with success that the new technologies can
  be solid building blocks for the LHC computing
  environment