Design of a Model-Based SLA Management Service and Its Applications over Distributed SOA Environments - Grid Computing: Infrastructure, Service, and Applications - page 249

10.2
Negotiation has been widely studied in recent decades, in particular
human negotiation (e.g., negotiation between a seller and a buyer). SLA
negotiation and specii cations [8,10,13] for computation services has
recently attracted much interest as it points out the value of economic-
driven computer service provision. There are a number of related stud-
ies in SLA management that look at automation of service negotiation
[8,9] to provide network-based services according to established ser-
vice-level management. Literature [14,15] studies the efi ciency of SLA
negotiation processes and suggests ways of achieving mutually satis-
factory agreement with shorter renegotiation rounds/times and higher
probabilities of the acceptance. In [14], analytical approaches for nego-
tiation support systems (NSS) are proposed to optimize the outcome for
negotiation parties. WS-agreement was proposed by the GGF [16] to
describe the SLA-relevant data in a structured way. WS-agreement [9]
proposes an XML-based language to specify SLA templates that the ser-
vice providers can issue to their clients for i lling their requirements. It
also suggests a generic way of constructing the SLAs, SLA proposals,
and offers in XML . Methods in [15] suggest extending WS-agreement
by specifying the conditions of the services in guarantee terms so as to
improve the chances of reaching an agreement and reduce negotiation
trafi c.
In a distributed cooperative infrastructure, SLA management [17] is not
only the gateway between the service provider and its customer for the
service negotiation, but also the interface for services to interact with each
other. For example, an application service provider may negotiate with a
resource service provider to ask for the physical resources. SLA decompo-
sition [18] is an important technique for translating the QoS requirements
of the application into the requirements of the physical resources . However,
the difi culty of accurately predicting the performance motivates the
study of adaptation and optimization in SLA management. Literature [19]
suggests establishing SLAs with various QoS levels where each QoS level
is associated with service price and violation penalty. Heuristic algorithms
are proposed for the service provider to admit the requests selectively and
adapt the services according to the runtime resource availability so as to
maximize its global proi ts.
The purpose of SLA-driven resource management [20] is to manage
the distributed resources to ensure service quality and this has been
studied in recent research work [21-24]. Literature (e.g., GRAM [21],
GARA [22], SNAP [23], VAS [24]) considers ensuring the service
quality by reserving resources in advance; for example, admitting/
rejecting requests, choosing appropriate resources, or reserving and

Related Work