A Model-Based Autonomous Engine for Application Runtime Environment
Configuration and Deployment in PaaS Cloud
XingTu Lan
1,2
, YanPing Liu
1,2
, Xing Chen
1,2
*, YiCheng Huang
1,2
, Bing Lin
1,2
, Wenzhong Guo
1,2
1 College of Mathematics and Computer Science, Fuzhou University, Fuzhou 350108, China
2 Fujian Provincial Key Laboratory of Networking Computing and Intelligent Information Processing
fzulanxingtu@gmail.com; fzulyp@gmail.com; chenxing@fzu.edu.cn; michael.huang.aiesec.fzsu@gmail.com;
WheelLX@163.com; guowenzhong@fzu.edu.cn;
Abstract—Cloud Computing is evolving as a key computing
paradigm for sharing resources. One type of Cloud, which
provides platform resources including all the elements of
application runtime environment, is regarded as PaaS Cloud.
The management of PaaS Cloud is a complex task, up to the
point, where manual operation is hard to be cost effective. As
the application runtime environment is supported by a set of
dynamically composed, distributed elements. What is more, in
order to achieve a management target, multiple operations
have to be applied over the distributed and heterogeneous
elements of PaaS Cloud. To improve the management of PaaS
Cloud, this paper proposes to support the configuration and
deployment of application runtime environment in PaaS Cloud
with an autonomous engine. The automation is enabled by the
definition of a domain-specific information model, which
captures all the related information with the same
abstractions, describing the application runtime environment,
PaaS Cloud infrastructure and management targets. On top of
that, a technique based on Satisfiability is described, which
automatically analyses the state of the managed objects and
plans required operations for maintaining it. The result from a
case study is provided to validate the feasibility of this
approach.
Keywords-Cloud Computing; Platform as a Service; Model-
Based Management; Autonomic Systems; Satisfiability.
I. INTRODUCTION
Cloud Computing is evolving as a key computing
paradigm for sharing resources. One type of Cloud, which
provides platform resources including all the elements of
application runtime environment, is regarded as PaaS Cloud.
There are a lot of practical products of PaaS Cloud at
present. For example, many individual users deploy their
applications onto the public PaaS Cloud such as Google
App Engine [1] and Windows Azure [2]. Many enterprises
also built their own private PaaS Cloud, which are used to
support some kinds of enterprise applications. Nowadays,
PaaS Cloud seems to be preferred way to implement the
computing platform in industry, as resources could be
allocated dynamically.
There are multiple challenges of PaaS Cloud
management to fit to business requirements. First,
application runtime environment contains different levels of
software and hardware infrastructure. In each level, different
types of products exist. Second, both PaaS Cloud and
application runtime environment are supported by a set of
distributed elements. Third, exiting constraints of
application runtime environment should be taken care of.
Fourth, PaaS Cloud supports many pieces of runtime
environment and it should meet the business requirements
itself. In order to achieve a management objective, lots of
operations have to be applied over the distributed and
heterogeneous elements of PaaS Cloud. At present the
factors of heterogeneity and distribution have been taken
into account by operation tools following the traditional
management paradigm – an administrator working with a
management console. However, reasoning activities such as
adaptation and decision making on the information provided
by PaaS Cloud are seldom automated, resulting in a true
bottleneck for the management.
In this article we present our approach to improving the
automation capabilities of the configuration and deployment
of application runtime environment in PaaS Cloud. The
main contribution of this paper is the characterization of a
domain-specific information model of PaaS Cloud which
captures all the related information with the same
abstractions, describing the application runtime
environment, PaaS Cloud infrastructure and management
targets. Building on top of that, we have designed a
technique based on the Satisfiability (SAT) [3], which
automatically analyses the state of the managed objects and
plans required operations for maintaining it.
The article is structured as follows. Section 2 gives an
overview of our approach. The next one describes the
information models we use in our solution and proposes
some key management definitions built over them. Section
4 details the algorithms for reasoning upon them
autonomously and the resolver to reach a desirable and
stable configuration. In Section 5 the proposal is validated
through a case study derived from a real scene. Finally, the
last section reflects on the explained aspects, focusing on
the impact and potential extensions to this work.
II. O
VERVIEW OF APPROACH
The management approach for PaaS Cloud that we are
proposing mainly rests on two principles. On one hand, the
management system adopts a standard information model to
capture all the relevant information about the managed
environment in a generic way. On the other hand, the
management in PaaS Cloud should be valid and effective. In
2014 IEEE 6th International Conference on Cloud Computing Technology and Science
978-1-4799-4093-6/14 $31.00 © 2014 IEEE
DOI 10.1109/CloudCom.2014.80
823
2014 IEEE 6th International Conference on Cloud Computing Technology and Science
978-1-4799-4093-6/14 $31.00 © 2014 IEEE
DOI 10.1109/CloudCom.2014.80
823