Why? Local batch schedulers use a great deal of
information to share out resources amongst competing users and maintain high levels of resource usage.
Grid scheduling is very simplistic, routing jobs to a site with an idle CPU (perhaps with the right data available).
How do we bridge the gap?
Question by LHCC member at gLite demo, Tuesday:– “If there is an urgent analysis, how does gLite send
this to a special job queue?”
Sessions Thursday AM
– Description of how different local schedulers are used at HEP sites.
– What are the common queue management features? Thursday PM
– What is the impact of today’s simplistic Grid scheduling?
– How do local and Grid-level scheduler developers see the future?
Friday– Discussion
» Can/how do we move towards better Grid scheduling?» Can we define common HEP environment variables to ease
transfer of jobs between sites?