Information Technology Reference
In-Depth Information
PIndex requires a highly dynamic structure, making most grid simula-
tors unsuitable. A more specii c simulator is needed, which can be tai-
lored to model a dynamic environment with many thousands of nodes.
Although P2P network simulators [18,19] are available, they only work
for specii c P2P network structures, which cannot be easily modii ed
for PIndex.
13.4.2
Choosing Colored Petri Nets for PIndex Modeling
A Petri net is a modeling language that can be used to graphically depict
the structure of a distributed system. Consisting of two main elements,
places and transitions, many complex processes can be modeled. More
importantly, it enables one to easily depict parallel processes. A Petri net
has place nodes, transition nodes, and directed arcs connecting places
with transitions. Petri nets have tokens that are represented as dots indi-
cating the occupation of a place caused by a previous transition. These
tokens either occupy a place or not (akin to binary 0, 1), leading to the
representation of Petri nets as matrices.
Much work has been proposed to extend the original Petri net concepts.
For example, stochastic Petri nets (SPNs) [20] allow timings to be imple-
mented to Petri nets giving more realistic complex models, and include
the abilit y to invoke models of probabilit y distributions of events. Another
concept is colored Petri nets (CPNs) [21], which allow tokens with distinct
characteristics to be represented by colors simplifying the depiction of a
complex Petri net for various functions. The following main points have
been made to further demonstrate the rationale in choosing Petri nets to
model PIndex.
•
The nodes in PIndex may have many different states. These can
be depicted with colored tokens.
The primary concern of PIndex is to perform through the use of
•
messages; this provides a level of coarse granularity, which can be
represented by transitions in Petri nets.
As many PGs operating in parallel and interact with each other in
•
PIndex, the ability to model concurrent events is needed. This
requirement is intrinsic to Petri nets as they were designed with
concurrency in mind.
Since PIndex is a robust and fault-tolerant network, these features
•
must be able to be simulated at any time. Petri nets have the abil-
ity to change states during a transition, allowing for failures to
occur randomly.
Each node in PIndex must operate independently, which can be
•
mimicked by the use of tokens as objects in CPNs.
Search WWH ::
Custom Search