CMSLCG0 architecture

1
CMS/LCG-0 architecture

• Many authors

2
CMS/LCG-N testbed

• CMS/LCG-0 is a CMS-wide testbed based on the LCG
  pilot distribution (LCG-0), owned by CMS
• Red Hat 7.3 (7.3.2 with CERN kernel recommended)
• Components from VDT 1.1.6 and EDG 1.4.X (LCG
  pilot)
• Components from DataTAG (GLUE schemas and info
  providers)
• Virtual Organization Management VOMS
• RLS in place of the replica catalogue (uses
  rlscms by CERN/IT!)
• Monitoring GridICE by DataTAG
• R-GMA (as BOSS transport layer for specific
  tests)
• Dynamic installtest new components useful to
  CMS
• Support (outside CMS)
• DataTAG WP4
• LCG Experiment Integration and Support (EIS) team
• Currently configured as a CMS RC and producing
  data for PCP
• It allows us to do our software integration while
  LCG-1 is not out!!!

3
CMS/LCG-0 sites
170 CPUs
4
CMS/LCG-0 as a CMS Regional Center
Users Site (or grid UI)
Resources
Production Manager defines assignments
shell scripts
RefDB
Phys.Group asks for an official dataset
Local Batch Manager
CMS LCG-0
JDL
LCG Scheduler
Site Manager starts an assignment
MCRunjob ImpalaLite or CMSProd
USCMS testbed
DAGMan (MOP)
User starts a private production
Chimera VDL
Virtual Data Catalogue
Planner
5
CMS/LCG-0 PCP setup
SE
Job management
BOSS DB
RefDB
UI CMSProd/ MCRunbjob BOSS
SE
Workload Management System
input data location
SE
CE
cmsrls
SE
Data replication
SE
BOSS DB
CERN SE
SE
UI
cmsrls
SE
6
Outcomes

• 4 UIs testing different production tools
• McRunjob/ImpalaLite (Bari, Bologna)
• complex! But consistent with official
  environment
• CMSProd (Padova, Ecole Poly.)
• simple!
• Testbed is quite stable
• 500 Kevts heavy CMKIN (8h/job(250 evts) 20
  GB)
• 500 Kevts CMSIM in production (350 Kevts/week
  320 GB)
• Operates as a CMS Regional Centers
• need to interact with existing environment, e.g.
• not grid-aware comm. to/from RefDB e-mail
• not grid-aware data transfer tools SRB bridge at
  CERN
• much more consistent architecture if we had only
  grid components!
• Only production services (no user analysis)