Information Technology Reference
In-Depth Information
1.
informed(
,T,s,p
)
Creates an initial partitioning or updates an existing partitioning.
The function takes the set of agents
A
, the set of available threads
T
,
the partitioning strategy
s
, and an eventually existing partitioning
p
as input. It returns a new partitioning.
A
2.
updateNecessary(
p
)
Evaluates whether the current partitioning
p
has to be updated or
not.
Additional information is required to evaluate whether an update
of a partitioning is necessary or not, and to actually compute an
updated partitioning. Collection and computation of this (statistical)
information are not included in the functions mentioned above, but
will be described in context of specific partitioning strategies.
Besides these two functions, the constraint evaluation has to be
adapted to make use of the additional information. Each partitioning
strategy will therefore require a specific adapted function for evaluating
constraints.
Spatial decomposition
Spatial decomposition aims at reducing the number of evaluations
whether two or more actions are interfering by spatial separation. The
improvement gained by this separation is that only actions within the
same spatial region need to be checked for interference.
Known approaches include the
Spheres of Influence
proposed by
Logan and Theodoropolous [78, 127] and the notion of
Event Ho-
rizons
by Scheutz and Schermerhorn [114]. Steinman also follows
this approach and proposes a so-called 'grid-based publish and sub-
scribe data distribution system' [125]. This approach is similar to
using well-known quad-trees or oct-trees for spatial representation of
a model. One challenge is to determine the size of the grid which
is equivalent to determining number and layout of spatial regions
respectively partitions [125].
