Information Technology Reference
In-Depth Information
the service client to sign. The client will i rst review the SLA proposal2
created by the service provider before signing it. The SLA proposal2 will
i nally become an SLA as the signatures from both the service provider
and the client are present.
In real-life situations, there may be other possible SLA negotiation
sequences determined by who initiates the SLA negotiations, who signs
the SLA proposal i rst, whether the service provider or/and the client
will give a counter-proposal by modifying the previous SLA proposal, and
so on. The SLA copy management in our proposed SLA-based service is
l exible and extensible to deal with various scenarios of SLA negotiation in
distributed environments.
10.5
In this section, we will examine the performance of our proposed SLA
manager by incorporating it with Fire Dynamic Simulation (FDS) for pro-
viding high-performance service over the shared services platform.
Applications of Fire Dynamic Simulation Services
10.5.1
Deployment for FDS Services with an SLA Manager
FDS is a computer program based on computational l uid dynamics (CFD)
models to simulate i re growth and spread. “The software solves numeri-
cally a form of the Navier-Stokes equations appropriate for low-speed,
thermally-driven l ow, with an emphasis on smoke and heat transport
from i res” [26]. FDS was developed and maintained by the National
Institute of Standards and Technology (NIST) from 2000, and it was widely
used to simulate i reworks and predict the temperature and radiation of
heat and smoke. The FDS works with Smoke View, a visualization program,
to demonstrate the propagation of the i re and estimate the damages it
may cause within a certain amount of time.
As a high-performance application, FDS generates a large amount of
data and consumes signii cant computation resources. It may take extra-
ordinary processing time to run a simulation test especially when there
are no sufi cient resources available. In general the processing time of FDS
will be affected by resource availability, environmental data setting, and
simulation parameters such as simulation duration, number of CPUs, and
so on. For example, one FDS simulation test (e.g., 3st-5grid) with 1-minute
simulation duration on i ve CPUs takes about 30 minutes to complete the
job running. FDS is programmed into two different modes: sequential and
parallel. In a sequential mode, the simulation is done on one processor
only. In a parallel mode, the simulations are divided into multiple meshes
that can run in parallel on multiple processors using MPI. Users can
Search WWH ::
Custom Search