Information Technology Reference
In-Depth Information
the infrastructure. At the same time, enterprises, especially IT companies, tend to set
up PaaS platforms to provide Application Execution Environment (AEE) for their
applications and reuse a variety of application components. Because PaaS also has to
rely on the infrastructure such as servers and storage, IaaS becomes a candidate to
provide PaaS the infrastructure. That is the PaaS based on IaaS (PoI). PoI could pro-
vide the features of both PaaS and IaaS: i) to provide AEE for applications and appli-
cation components as PaaS; ii) to increase the utilization rate of the infrastructure
as IaaS.
Resource management is one of the most important issues of cloud computing which
decides the efficiency of the platform. In PoI, resource management also plays a very
important role as well. It has many characteristics: i) the traffic model and QoS require-
ments of different applications on PoI are different and changing fast. ii) PaaS could
dynamically use the resource of IaaS according to the requirements of applications.
In this paper, a QoS guaranteed and cost-efficient resource management mechan-
ism in PoI is presented. The rest of the paper is organized as follows. The related
works are introduced in Section 2. Then the architecture of PoI is illustrated in
Section 3. In section 4, the problems which have to be faced by PoI are stated and the
resource management framework of PoI is brought forward in detail. In the end,
the
experimental results are presented in Section 5 followed by the conclusions and future
works in section 6.
2
Related Works
Recent research on dynamic resource provisioning in virtualized environments in-
cludes [1]. These research attempted to achieve application performance goals with
dynamic resource provisioning in virtualized environments. Most recently, people
extended the idea into the cloud environment [2]. [3] investigated task scheduling
with deadline and budget constraints in the heterogeneous environment. Some re-
search investigated cost-efficiency in the cloud environment. On the cloud providers'
point of view, [4][5] discussed the resource allocation and instance consolidation
strategies for cloud data centers. The goal is to maximize the profits of cloud provid-
ers while maintaining Service Level Agreement (SLAs). [6] discussed power man-
agement in cloud environment. On the cloud users' point of view, the hot topic is to
build strategies to deploy applications among multiple cloud providers to enhance
availability with minimum cost [7].
Because PaaS platform is essentially a distributed system, it faces similar resource
management problems like other distributed systems. Computing resources (e.g.,
CPU, memory, disk I/O and network bandwidth) are limited in distributed systems;
therefore how to manage these resources with high efficiency is critical. PaaS plat-
form is asynchronous soft real-time system [8]. On one hand, PaaS platform is
asynchronous because requests of applications on PaaS are unpredictable and nonde-
terministic, and the distribution of the requests cannot be precisely described by ma-
thematical models. On the other hand, it is a soft real-time system because there are
some
constraints on the response time, but the failure of timeliness will only lead to
