Proposal for a IS schema

1
Proposal for a IS schema

• Massimo Sgaravatto
• INFN Padova

2
Introduction

• We have to converge on a set of information
  needed by our scheduler (for M 9 release)
• We have to agree with the other WPs on a common
  schema
• We have to negotiate with other WPs (WP4, ) who
  is in charge to implement the required
  information providers
• As main customer of the monitoring and
  information service we should make a proposal

3
Proposal

• Draft proposal for a new schema proposed in the
  context of WP 1 INFN-GRID (Evaluation of the
  Globus toolkit)
• http//www.infn.it/globus/Docs/gis-farm.pdf
• Useful for our needs ???

4
Assumptions (for M9)

• Globus based prototype
• A computing element is represented by a set of
  local queues of an underlying resource management
  system
• The Grid Scheduler submit jobs to queues
• A queue represents a set of homogeneous
  resources
• When a job is submitted to a specific queue, it
  doesnt matter in which node of this queue the
  job is dispatched

5
Proposed schema
6
Jobmanager service (front-end to the farm)

• Host name of the front-end machine
• Name of job manager (jobmanager-lsf,
  jobmanager-pbs, )
• Contact string
• Resource management (LSF/PBS/)
• Which local resource management systems do we
  have to consider at M9 ?
• Resource management version
• GRAM Version

7
Queue

• Name of the queue
• Architecture
• Operating System
• Total CPUs
• Free CPUs
• Total number of jobs (running and pending) in the
  queue
• Number of running jobs
• Number of idle jobs
• Maximum number of running jobs allowed for this
  queue
• Maximum number of jobs (running and idle) allowed
  for this queue
• Status of the queue (ready or not ready to
  dispatch jobs to the executing machine)
• Run windows (when the queue is active)
• Priority
• Maximum CPU time for a job submitted to this
  queue
• Maximum wall clock time for a job submitted to
  this queue

8
Job

• Grid job ID
• Local job ID (LSF job id, Condor job id)
• Grid user id (subject of his/her certificate)
• Local user id
• Pending reason
• JDL expression
• Submit time
• Start time
• Wall clock time

9
Missing info

• What about the authorization policies ?
• Grid-mapfiles ?
• What about the scratch area ?
• Available disk space
• ?
• What about the storage elements ?
• What about the application environments ?
• ???