Information Technology Reference
In-Depth Information
Currently, with the energy costs increasing, the focus shifts from optimizing
Data Center resource management for pure performance to optimize it for en-
ergy eciency while maintaining the level of services [2].
In a Cloud Computing environment, there are distinct participants with dis-
tinct objectives, preferences and disposition to pay for services. In this scenario, a
Multi-agent System (MAS) can be used where each participant is an autonomous
agent that incorporates market and negotiation capabilities.
In this work, we propose a Federated Application Provision (FAP) strategy
which uses multiple agents and server consolidation techniques to achieve power-
aware resource allocation, by taking into account SLAs, energy consumption and
carbon footprint. In our approach, the user should pay according to the eciency
of his/her applications in terms of resource utilization and power consumption.
Therefore, we propose that the price paid by the users should increase according
to the whole energy consumption of the Data Center, especially when the user
does not accept to negotiate QoS requirements.
Experimental results for our FAP consolidation strategy were obtained in the
CloudSim [3] simulator, with 2 Data Centers, each one belonging to a different
Cloud, and 400 simultaneous virtual machines show that our approach is able
to reduce an average of 53.57% of energy consumption, while meeting the SLA
requirements.
The remainder of this paper is organized as follows. Section 2 presents concepts
of Cloud Computing. Section 3 discusses energy green performance indicators.
The proposed strategy for Federated Cloud server consolidation is presented in
Section 4. In Section 5, experimental results are discussed. Section 6 presents
related work. Finally, Section 7 presents the conclusion and future work.
2 Cloud Computing
There are many definitions of cloud computing in literature. Most of these defini-
tions state that a Cloud Computing system should have (i) pay-per-use capabil-
ity, (ii) elastic capacity and the illusion of infinite resources, (iii) self-service, (iv)
virtualized resources and (v) QoS enhancement functionality. The cloud service
models are divided in three classes, according to the abstraction level and the
service model of the providers: Infrastructure-as-a-Service (IaaS), Plataform-as-
a-Service (PaaS), and Software-as-a-Services (SaaS) [17].
In the Infrastructure-as-a-Service (IaaS) model, the user can request process-
ing power, storage, network and other fundamental computing resources such
as the operating system, for a period of time and pay only what he/she uses.
Plataform-as-a-Service (PaaS) are development platforms that allow the creation
of applications with supported programming languages and tools hosted in the
cloud and accessed through a browser. This model can slash development time,
offering readily available tools and services. In the Software-as-a-Service (SaaS)
model, applications run on the Cloud infrastructure and are accessible from var-
ious client devices. From the user view, the SaaS model allows him/her to save
money in servers and software licenses.
