Information Technology Reference
In-Depth Information
resource requests, but very few take into account KPIs such as the business value
of the request.
The SLA
evaluation
capability compares information collected from sensors
and other monitoring tools to the SLA objectives, and raises alarms when thresholds
are passed or constraints are violated. The provider, having detailed information of
its resource status, can act proactively to address failures, thus managing the risk
associated with the penalties incurred. Depending on the SLA, the consumer may
also receive such notifications, and can reallocate tasks, enhancing its ability to
react to the likelihood of failures. This capability builds on a modular architecture
that exploits a topics-oriented publish-subscribe model and can instrument native
or “off-the-self” ICT resource monitoring tools. Intelligent event correlation and
non-repudiation combined with SLA-based accounting and annotation of violations
make information clearer and reliable enough for enabling evidence gathering and
evidence-based decision making regarding claims for compensation.
The
SLA-based accounting
capability supports the selection and adoption of
the suitable charging scheme for each service execution environment based on the
metrics included in the SLA specification. Such metrics may deal with a variety of
heterogeneous resources. This capability enables charging for service use based on
its real execution cost. The analysis reports produced by this capability also helps
clarifying resource usage and causality of retribution and penalties.
Figure 8.7 shows an overview of these capabilities against the typical life-cycle
of managing SLAs. The business benefits of such a comprehensive environment
for SLA management over Grid middleware include optimising resource alloca-
tion and use in response to market requirements, reducing Total Cost of Ownership
(TCO) by improving efficiency of resource utilisation and faster and better targeted
response to failures, increasing customer confidence by allowing transparency of
operation (subject to the SLA), and enabling customisable billing by providing
finer granularity of accounting and reporting. These results have been validated
in various Business Experiments in different vertical market sectors including the
Business Experiments in the area of online collaborative gaming (see chapter 12)
and remote computation for radiotherapy cancer treatment planning (see chapter 9).
8.6 Common Capabilities for Data Management
Companies in most vertical market sectors that are considering the use of Cloud
computing or Data-Grids for federating data share common concerns about storage,
access, translation and integration. These can be simplified in the following key
points - further analyses are included in Dimitrakos et al. (2009a) and Thomson
(2009):
• Where should data be placed and how should it be retained?
• How should data be accessed?
• How should data be presented by one provider so that others will understand it?
• How can one combine data from many distributed and heterogeneous sources?
