[ogsa-wg] Workflow use-case

Donal K. Fellows donal.k.fellows at manchester.ac.uk
Mon Apr 16 04:30:55 CDT 2007


Hi everyone

I've been meaning to write this up for a while now. This is a use-case
for workflows that is derived from one of the key use-cases in the EU
BREIN project that I'm working on. The scenario is that there is a need
to perform a complex sweep over a non-trivial parameter space in a
limited period of time as part of a series of in-silico experiments: an
example might be testing compositions of and configurations of turbine
blades through demanding duty cycles. To realize this, there is a need
for an outer engine that performs the parameter sweep and an inner part
that does data management[*] and which then runs some coupled
simulations on the dataset, quite probably with additional stages as
well to set up things like finite element meshes and stuff like that
(we're interested in doing coupled CFD and FEM codes; turbine blade
design is a good example.) There's a desire to keep the amount of work
done down to keep turnaround time short, and there's also a desire to
manage things in such a way that if a component fails[**], it remains
possible to recover and still complete the overall workflow goals. We
specifically want to be able to support a workflow system that can
manage multiple ways of achieving the overall goal or a subgoal; I for
one want to run a test where I pull the plug on one of the servers in
use (i.e. a total catastrophic failure!) and have the customer's overall
SLA still met.

I don't know if I've written up enough details yet.

Donal.
[* Our specific scenarios have the data moved to the computation for
    security and legal reasons. ]
[** There are a _lot_ of failure modes! Many of them are recoverable
     with the right sort of workflow manager, but maybe not all. ]


More information about the ogsa-wg mailing list