Information Technology Reference
In-Depth Information
federates or RTIExec processes on each site, such as saving and restoring
a federate for migration. Broker support services are dei ned and they
reside on each federate site providing necessary information about the
local federate such as the current host load and application behavior. The
migration service is dei ned to coordinate federate migration between
source and destination sites by communicating with their interfacing ser-
vices. To decide when and where federate migration should be performed,
collective services are dei ned including the application monitoring main
service manager, the performance decision service and the broker service.
They communicate with the broker support services of each federate for
global performance monitoring, decision-making, and migration trigger-
ing. Although the various grid services cooperate for an optimized con-
i guration of federation execution, the actual simulation communications
are through a vendor-specii c RTI.
Wu et al. [10] proposed a data grid called Aegis for large-scale distrib-
uted simulation applications. It offers data resource management services
and computing services for HLA-based simulation applications while the
simulation execution still uses a vendor-specii c RTI. Choi et al. [11] pro-
posed an RTI execution environment called RTI-G based on the OGSA. In
order to achieve high performance, RTI-G utilizes various GT3 services,
such as MDS, GRAM, and GridFTP, to enable dynamic resource allocation
and automatic execution of HLA-based distributed simulations.
A simulation grid is an infrastructure that integrates grid technology
and the HLA to realize dynamic and secure resource sharing, optimized
resource utilization, collaborative activities and fault tolerance for dis-
tributed simulation applications [12,13]. In a simulation grid, HLA-based
distributed simulations are grid-facilitated in various ways. For example,
registration services are dei ned for dynamic model resource discovery
and scheduling services are dei ned to generate proper simulation deploy-
ment schemes according to monitored computing resource status.
Our group proposed a load-management system for HLA-based dis-
tributed simulations in [14]. The load-management system and its federate
migration mechanism are based on grid services while communications
between federates are through a vendor-specii c RTI. It was later extended
with an efi cient migration protocol in [15].
We also proposed a grid service-based framework for l exible execution
of large-scale HLA-based distributed simulations [16]. RTI executive pro-
cesses and federate models are encapsulated in grid services for dynamic
federation creation and management. An index service is provided for
registration and dynamic discovery of the various grid services. Simulation
communications are also through a vendor-specii c RTI.
As the grid-facilitated approach still replies on a vendor-specii c RTI for
simulation communication, it requires cross-domain trust and particular
prior security setup for HLA-based distributed simulations to be con-
ducted across administrative domains, which is very cumbersome.
Search WWH ::
Custom Search