Speaker
Grehant Xavier
(CERN openlab / HP)
Description
The variety of EGEE applications come up with many different
requirements. For
example, biomedical and generic virtual organizations (VOs)
run jobs for five
minutes, whereas high energy physics (HEP) jobs can run for
days. Some mechanisms
have been investigated to make this cohabitation
satisfactory in leveraging theories
from economics. Additionally, thanks to virtualization, it
is now possible to make
the allocation as dynamic as electric power supplies, and to
implement schemes that
lead to an optimal resource utility. Such a system has been
implemented in HP Labs
and tested in CERN openlab. An interesting approach in the
future would be to have
our grid infrastructure benefit from it without losing the
capabilities of current
allocation architecture and strategies, like static
contracts between VOs and
service-level agreements. If successful, such an integration
could also be an early
step towards organizing grid self-sustainability.
Author
Grehant Xavier
(CERN openlab / HP)
