1. SA1 ROC Meeting
Bologna, October 2004
Priorities for grid operations Bob Jones EGEE Technical Director
EGEE is a project co-funded by the European Commission under contract INFSO-RI

2. Priorities for grid operations
Succeed in passing the 1st EU review (Feb ) by demonstrating that we have established a large-scale production quality grid infrastructure supporting multiple applications and user communities
The active involvement of the ROCs is essential to meet this goal
From Tech Annex: “In terms of funding and manpower, the ROCs represent the single biggest item in the EGEE proposal”
ROC managers, Bologna, October

3. Specific goals for grid operations
Deployment and support of non-HEP applications
Continued support for HEP data challenges
Complete migration from LCG to EGEE operations structure
Continued expansion of the production service
Deployment of pre-production service providing early access to gLite components for applications
EU review will include live demonstrations of HEP and BioMedical applications running on the production service. gLite will also be demonstrated.
ROC managers, Bologna, October

4. Production Service Resources
Federation
TA: Month 1
TA: Month 15
LCG-2: Actual
delta Month 1
CERN
900
1800
956
56.00
UK&I
100
2200
2132 Fr
400
895
160 It
553
679
1836 SE
146
322
108
-38.00 SW
250
408
158.00 CE
385
730
210 NE
200
2000
348
148.00
DE+CH
910
810.00 Ru 50
152
169
119.00
totals
3084
9428
7179
Others (Taiwan, China, Canada, US, Pakistan & India)
~2000
ROC managers, Bologna, October

5. Resources for non-HEP applications
Resources needed for non-HEP applications thru end 2004:
BioMedical:
GATE
100 CPUs, several tens of GB storage and access to a supercomputer (CINES Montpellier)
NPSA (
CPUs: 10 sites with 10 CPUs each and short queues. IO file size:  from kB to GB. file storage: 500 GB
Drug discovery
CPU usage: 4 150 h, input file size: 10 MB, output file size: 1-2 MB, number of jobs: 
Generic:
Earth Observation
Storage Elements: 500GByte, distributed over several Centers
Computing Elements: Cluster with fast connection like Mirinet to run IPGP MPI jobs
Total computer time predicted to be used is around hours
Astroparticle physics
200 CPUs and 5TByte of storage
Earth Science
Up to 100 CPUs and 100Gbyte of storage
Computational chemistry
128 nodes + 16 CPUs with at least 2GB of RAM
Provision of missing month 1 resources
would cover most of these requests
ROC managers, Bologna, October

6. Resource Negotiation Via the PMB, the federations have been asked to:
Provide a statement about national policy for the allocation of resources.
Example: “African federation: in Nigeria we will provide resources connected to EGEE for applications from HEP (90%), Biomedical (5%), computational chemistry (3%), earth observation (2%). In Benin we will provide resources for applications from Biomedical (50%), astrophysics (30%), geophysics (20%), etc.”
Empower ROC managers to use this information as a guideline when negotiating resource allocation with participating sites
ROC managers, Bologna, October

7. Relationship with CICs
Proposed CIC agreement
Each of the CICs (RAL, CNAF, Lyon, CERN) assigns people to operations management, with responsibility cycling round the sites on a weekly basis
daily run of the validation test suites
monitoring the roll-out list, filtering out problems
overview of the general availability and performance monitors
based on the results of the above - generation of trouble tickets
lead the investigation of difficult problems
active management/escalation of trouble tickets
intervention as necessary with ROCs/primary sites
assembly of FAQs, recipes and other useful documentation
proposing and developing additional tests, automating processes, etc.
The CICs would normally communicate with the ROC/Tier-1/Primary Site, which is in turn responsible for taking action directly with the site.
Where there is no appropriate ROC or Tier-1 a "Primary Site" will be assigned (usually one per country)
ROC managers, Bologna, October