This work was partially supported by the National Natural Science Foundation of China (No. 71161015, 61462056,61462051, 81360230), the Applied
Fundamental Research Project of Yunnan Province (No. 2014FA028, 2014FB133, 2013FA013, 2013FA032).
A Method of QoS Deviation Monitoring for Web Services
CHEN Bin-Bin
1
, FU Xiao-Dong
1,2*
, LIU Li
1
, GAO Tian-Yu
1
, GE Jia
1
1. Faculty of Information Engineering and Automation, Kunming University of Science and Technology, Kunming 650500
2. Yunnan Provincial Key Laboratory of Computer Application, Kunming 650500
E-mail: xiaodong_fu@hotmail.com
Abstract: Considering the fraudulent behavior of Quality of Service (QoS) for web services, a method to monitor the QoS deviation was
proposed in this paper. Firstly, we used random variable to describe the uncertainty of QoS. Secondly, the divergence of QoS in different time
periods was tested with Kolmogorov Smirnov (K-S) approach. Finally, we utilized stochastic dominance to judge the better one of them.
Theoretical analysis and experimental data showed that the proposed method was reasonable and effective.
Key Words: Quality of Service, Web Services, Kolmogorov Smirnov, Stochastic Dominance
1 INTRODUCTION
Web service is a service-oriented architecture (SOA)
distributed computing technology. In essence, web service is a
self-description and modular application identified by URL. It
uses the open standards based on XML and the Internet. By
dynamically discovering the already existing web services and
organically composing them with certain rules, they can be
assembled into a valuable and higher-granularity composite
service which has internal logic to meet the complex needs for
different users and achieve the business objectives. It has
become the main implementation techniques of service
computing [1].
The composed web services are published by different
organizations on the Internet, realizing all kinds of web
services integration and resource sharing in the open, dynamic
and difficultly controlled network environment. So the QoS
becomes one of the key factors to determine whether they can
succeed [2]. The QoS for web services include response time,
throughput, availability, reliability, and reputation, et al [3].
Because web services users do not own the software and
only care about the results, they pay more attention to the
perceived QoS. As the important indicator to evaluate web
services, the QoS is not only used for services selecting, but
also for guiding the development of high-quality composite
services [4]. However, because of existing too many
competitors with the same features, in order to improve the
opportunity of their services being selected, there may be
some fraudulent behavior between service providers. For
example, they may provide high-quality service in the initial
stage, but the QoS in the next phase diverge greatly from the
former. Finally, they announce that they have provided
high-quality web services in the whole period. In addition, the
complex environmental factors also make the QoS uncertain.
Due to the asymmetric information, service consumers can
not obtain the real QoS information which make them face the
risk of not really get their required services.
To this end, it is urgent for us to propose an effective
method which can determine whether exists the QoS
deviation at different time periods and monitor the deviation.
We do not make any assumption on the distributed form of the
QoS for web services. Firstly, we consider the two stage QoS
as two random variables and use the distribution function to
describe the random probability model. Then we can use K-S
test theory to judge whether there is deviation between the
two random variables. If you check it out, then you can utilize
the Stochastic Dominance theory to compare the merits of
them [5]. Finally, experiments verify the effectiveness and
practicality of this method.
1946
978-1-4799-7016-2/15/$31.00
c
2015 IEEE