Biomedical Engineering Reference
In-Depth Information
sweep” studies. Through the portal front-end,
a user can (1) register to the portal and create a
private space in the file system that will contain
the results of all the performed simulations; (2)
run a new simulation, after setting network and
protocol parameters, and graphically monitor the
performance indices and (3) choose a subset of
simulation data to graphically compare the related
performance results.
We hope that the SO-Grid portal will help
foster the understanding and use of swarm intel-
ligence, multi-agent and bio-inspired paradigms
in the field of distributed computing.
is placed according to a multicast mechanism
that aims to discover the data server which is the
closest to demanding clients.
The ARMAP protocol, proposed in this
chapter, is inspired by biological mechanisms,
and in particular exploits the self-organizing
and swarm intelligence characteristics of several
biological systems, ranging from ant colonies to
wasp swarms and bird flocks. In these systems, a
number of small and autonomous entities perform
very simple operations driven by local informa-
tion, but from the combination of such opera-
tions a complex and intelligent behavior emerges
(Dasgupta, 2005): for example, ants are able to
establish the shortest path towards a food source;
birds travel in large flocks and rapidly adapt their
movements to the ever changing characteristics
of the environment.
These biological systems can be quite natu-
rally emulated in a Grid through the multi-agent
paradigm (Sycara, 1998): the behavior of insects
and birds is imitated by mobile agents which
travel through the hosts of a Grid and perform
their operations. Multi-agent computer systems
can inherit interesting and highly beneficial
properties from their biological counterparts,
namely: (1) self-organization, since decisions
are based on local information, i.e., without any
central coordinator; (2) adaptivity, since agents
can flexibly react to the ever-changing environ-
ment; (3) stigmergy awareness (Grassè, 1959),
since agents are able to interact and cooperate
through the modifications of the environment
that are induced by their operations.
The ARMAP protocol is specifically inspired
to ant algorithms, a class of agent systems which
aim to solve very complex problems by imitating
the behavior of some species of ants (Bonabeau &
al., 1999). A technique based on ant pheromone is
exploited to tune the behavior of ARMAP agents,
making them able to autonomously switch from
the
copy
to the
move
mode. This kind of approach
is discussed in (Van Dyke & al., 2005), where a
decentralized scheme, inspired by insect phero-
bACkGROUND
The main purpose of the protocol presented in
this chapter is the dissemination of metadata
documents and their intelligent reorganization.
These two objectives are correlated, since an intel-
ligent dissemination of information can increase
efficiency management and facilitate discovery,
as discussed in (Forestiero & al., 2007).
Information dissemination is a fundamental
and frequently occurring problem in large, dy-
namic and distributed systems whose main pur-
pose is the management and delivery of content.
In (Iamnitchi & Foster, 2005) it is proposed to
disseminate information selectively to groups of
users with common interests, so that data is sent
only to where it is wanted. In our chapter, instead
of classifying users, the proposal is to exploit the
classification of resources: resource descriptors
are replicated and disseminated with the purpose
of creating regions of the network that are special-
ized in specific classes of resources. In (Aktas &
al., 2007) information dissemination is combined
with the issue of effective replica placement,
since the main interest is to place replicas in the
proximity of requesting clients by taking into ac-
count changing demand patterns. Specifically, a
metadata document is replicated if its demand is
higher than a defined threshold and each replica
Search WWH ::
Custom Search