Evaluation and Optimization of Mixed Redundancy
Strategy in Cloud-based Systems
He Pan
*
, Zhao Xueliang, Tan Chun, Zheng Zhihao, Yuan Yue
Chongqing Institute of Green and Intelligent Technology, Chinese Academy of Sciences, Chongqing 400714, P.R. China
Abstract: Mixed redundancy strategy is generally
used in cloud-based systems, with different node
switch mechanisms from traditional mixed strategy.
Existing studies on resources optimization often
concentrate on single strategy in cloud computing and
ignore the impact of mixed redundancy strategy.
Therefore, a model is proposed to evaluate and
optimize the reliability and performance of
cloud-based degraded system subject to mixed active
and cold standby redundancy strategy. In this strategy,
node switching is triggered by a continual monitoring
and detection mechanism once active nodes fail. A
continuous-time Markov chain is built on the state
transition process and numerical method is used to
solve the model and to compute transient availability
and expected job completion rate. After conducting
sensitivity analysis, a greedy search algorithm is
proposed to choose optimal redundancy for mixed
strategy under system constraints. Illustrative
examples were presented to explain the process of
calculating the transient probability of each state and
in turn, the availability and performance. It was shown
that near-optimal redundancy solution could be
obtained based on the analysis model. The comparison
with traditional mixed redundancy strategy proved
that the system behavior was different using different
kinds of mixed strategies and less redundancy was
assigned for the new type of mixed strategy.
Key words: Mixed redundancy strategy; monitoring;
reliability analysis; Markov chain; cloud-based system
I. INTRODUCTION
Cloud computing, which provides the hardware,
systems software or applications as services over the
Internet, enables the idea of computing as a utility [1].
Since cloud computing is often used for long-term
non-stop applications without the users’ control [2],
several types of software fault tolerance mechanism
are built in cloud-based systems, such as cold standby
redundancy [3], checkpoint-recovery [4], software
rejuvenation [2], failure prediction [5], system
monitoring [6,19], virtual machines (VM) migration
[7] and so on. Among these mechanisms, redundant
memory [8] or redundant nodes [6] is often
maintained in the system for VM migration. The
employment of redundant VMs often costs extra
resources and increases computing overhead. While
the energy consume has become a great concern in
cloud-based systems [9], the redundant resources
should be optimized. However, due to the distributed
features of cloud-based systems, multiple kinds of
strategies could be used for redundancy employment
in one system simultaneously [3]. Existing research on
resources optimization in cloud computing often
concentrates on single strategy and ignores the impact
of mixed redundancy strategies [9-11, 20].
Qiu et al. analyzed the failure rate and performance
of different kinds of software redundancy strategies
and selected strategies with respect to reliability and
resource constraint [3]. Nevertheless, the impact of
mixed strategy was not considered. Yang et al.
proposed a mixed redundancy mechanism for cloud
VM to improve the availability [6]. The availability of
the whole mechanism was not analyzed quantitatively,
so the number of both working and backup VMs was
only defined empirically. Mohamed and Al-Jaroodi
achieved dynamic load balancing and fault tolerance
through k-replica Cloud services and resources [12].
The reliability or performance was not analyzed and
how to choose the redundancy of each group of
services for the same partition of work was not stated.
In reliability engineering field, the analysis and
optimization of multiple redundancy strategies have
long been studied [13,14]. Statistical models are built for
reliability analysis and evolutionary algorithms are used
to search the optimal parameters. For multiple kinds of
redundancy strategies, Coit proposed the method to get
the optimal redundancy with different strategies used on
different subsystems and one strategy used in single
subsystem [15]. Ardakan and Hamadani conducted
redundancy allocation in series–parallel systems, in
which the active and cold standby redundancy was used
together in one subsystem [16]. In the above studies, the
cold standby redundancy was used only after all the
active redundant nodes failed. It was useful to extend
the operating time of a system but not suitable for
cloud-based systems in which performance was also
essential to job completion.
For the above problems, this paper proposes an
evaluation method to analyze the mixed redundancy
strategy in which cold standby VM is switched
immediately after active VM failure and to choose
appropriate VM redundancy under constraints.
Considering using both active and cold standby VM in
剩余8页未读，继续阅读
资源评论
