Now that we have a better understanding of the
conditions under which characteristics
of parallel programs can be used to reduce mean
response time and a better grasp of the size
of the reductions that can be expected,
we plan to
examine the importance of these job
characteristics and the relative performance of
the algorithms investigated here while considering
different job models.
(A modified model that includes limitations on
a job's maximum parallelism
as well as the maximum number of processors a job
can use efficiently has already been considered
[12].)
We hope to more fully explore variations on
our algorithm and to try to determine the
optimal allocation given the job, workload and
system models used in this paper.
We are also interested in trying to apply our
technique
in a static scheduling environment.