Information Technology Reference
In-Depth Information
Domain Ontology
completion time, mean time to failure,
mean time for recovery (tolerated annoy-
ance), application availability and associ-
ated cost. In some circumstances, the cor-
responding SLA values might differ from
values initially requested by the customer
because they may not reach.
It defines domain-specific knowledge so that the
SLA manager can have a semantic understanding
of a specific domain. This should be provided by
service provider and usually works with SLA.
Green Service
•
Customer Obligations: The customer need
commit to certain behavior in order to re-
ceive the required QoS at a given cost. For
example, the customer may simply be re-
quired to provide input data by a certain
time or may need to agree more complex
profiling such as using an application ac-
cording to certain usage time distribution
(mean usage rate for given time interval),
workload volume distribution, workload
complexity distribution. The customer
obligations will ultimately be recorded as
SLA terms. The process of deriving these
constraints can be quite complex and will
involve detailed knowledge to the custom-
er behavior.
Invoking a service involves the consumption
of resources and may incur cost. Sustainable
development advocates reduction of resource
consumption, while delivering better and more
widely available goods and services. Hence de-
livering “green service” has become increasingly
important. Recent concerns regarding global
climate change and the energy crisis have led to
renewed interest in Green Computing. In order to
address the issue, one of the aims of this proposed
QoS-oriented service computing methodology is
to investigate how we employ a variety of tech-
niques and tools to model the performance of
applications over a service-oriented infrastructure
for the trade-off among QoS guarantees, cost, and
margin profit to stakeholders in the value chain
involving service customer, service provider and
infrastructure provider.
The resource consumption of running an
application can be subject to several factors: (i)
application workload feature, (ii) user interac-
tion, (iii) network features, and (vi) mean time
to failure. We can use optimization technology
to find a resource with associated configuration
that can guarantee service's behavior within the
constraints and can maximize an objective func-
tion. In EU-funded IRMOS project (http://www.
irmosproject.eu/), the input parameters for the
optimization function have been indentified as
follows (Mitchell et al., 2009):
•
Application and Resource Profiles:
Application and resource profiles define a
set of parameters for applications and re-
sources that impact the key performance
indicators. The technical characteristics
of resource can include: specifications of
CPU, volatile storage, persistent storage,
operating system, system libraries; but
could also include scores of the platform
against a set of benchmark tests.
Application performance indicator predica-
tions can be carried out using a set of models
describing aspects of the user, the resource and the
application behavior, which are then combined to
determine the behavior of the service as a whole.
For example, in IRMOS project, it has identified
that each service can have the following models
(Addis et al., 2009):
•
Customer Requirements: The customer
requirements will be recorded in the SLA.
This will include key QoS performance
indicators such as application / service
Search WWH ::
Custom Search