Information Technology Reference
In-Depth Information
specify input parameters such as the number of processors, the value of
IBAR, JBAR, KBAR (i.e.,
x-
,
y-
,
z-
dimensional information) of a mesh, and
so on in an input data i le. If there are enough resources, all the meshes can
run in parallel; however, if the number of available processors is less than
the number of meshes, it will take a longer time to process.
We have implemented FDS as a service in SSP with SLA Manager as one
of the modules, as shown in Figure 10.8. Clients can subscribe to FDS
services via SLA Manager by specifying their job requirements. In the
meantime, the FDS service provider can negotiate with its potential clients
on the service terms and conditions based on client input parameters, esti-
mated simulation response time, resource availability, and so on.
The FDS service provider can not only establish SLAs with its service
clients, but also establish SLAs with other service providers to provide
comprehensive services for its clients. For example, the FDS service could
establish SLAs with other service providers that offer compute and stor-
age resources. As shown in Figure 10.8, the FDS establishes an SLA1, with
a physical resource service provider (PRsp) to obtain the computation
resources; for example, PC cluster, CPUs, and so on. In the meantime, an
SLA2, between the FDS service provider and the visualization service
Shared services platform
Visualization service provider
SLA2
Service
client
SLA2
SLA
Manager
FDS service provider
SLA3
SLA1
SLA3
FDS SLA
2/27/2008
SLA1
Input file
data.txt
Response time
<5 minutes
Physical resource service provider
No. of CPU
5
Storage space
1
GB
Service price
2.0
…...
FIGURE 10.8
SLA management of FDS service over SSP.
Search WWH ::
Custom Search