| description |
Ambient Intelligence (AmI) is an IT concept by which mobile users
shall be seamlessly supported in their every-day activities. This
includes interactions with remote resources as well as with their
current physical environment. We have developed the so-called Ad hoc
Service Grid (ASG) infrastructure that supports the latter form of
interactions. It allows operators to cover arbitrary locations with
Ambient Services in a Drop-and-Deploy fashion. An Ambient Service
may autonomously distribute (replicate and migrate) within an ASG
network to optimize its availability, response times, and network
usage. In this paper, we propose a fully decentralized, dynamic, and
adaptive service placement algorithm for AmI environments like the
ASG. This algorithm achieves a coordinated global placement pattern
that minimizes the communication costs without any central
controller. It does not even require additional communication among
the replicas. Moreover, placement patterns stabilize if no changes
occur in the environment while replicas still retain their ability
to adapt. Mechanisms for self-organized placement of services are
very important for AmI environments in general since they allow for
autonomous adaptations to dynamic changes and, thus, remove the need
for manual (re)configuration of a running system. We present a
detailed evaluation of the algorithm's performance and compare
it with three other algorithms to show its competitiveness.
Furthermore, we discuss how the desired self-organizing behavior
emerges from the interactions of a few simple, local rules that
govern the individual placement decisions. In order to do so, we
give an in-depth analysis of a series of hidden effects that are not
directly encoded into the placement algorithm but stem from its
collective dynamics.
|