基于Web服务质量因素的灰色关联分析中英文翻译

Grey Relational Analysis on Factors of the Quality of Web

Service

Abstract

Using the grey relational analysis theory, this exploratory study was to find the principal factors of Quality of Service of Web Sevice and its contributions. The paper utilized a real Quality of Web Service data sets: QWS Dataset (1.0) to grey relational analysis. The results indicate: (1) the sequence of influencing factors of Web Service selection is Best Practices, Compliance, Availability, Success ability, Reliability, Latency, Response Time, Throughput, Documentation; (2) nine factors can be classified into four kinds: the first important factors are Best Practices, Compliance, and Availability; the second important factors are Success ability, Reliability; the third important kind factors are Latency, Response Time; the less important factors are Throughput, Documentation; (3) the percent contributions of four kinds factors to the QoS of Web Service is 36.68% , 22.96% , 20.6% , 19.77%.

Keywords: Web Service; Quality of Service; Grey Theory; Grey Relational Analysis

1 Introduction

Web Service based on Service-Oriented Architecture[1,2] is an emerging distributed computing technology, which is gived more and more attention by industry and academia. With the development of Web Service technology and the widespread application, there are abundant Web Services in the Internet[3-5],Web Services searching merely based on the the functional requirements is unable to provide expecting service to the user. Quality of Service of Web Services plays the same role as functional attributes. How to select the same function services for the user is worth of research subject. Relevant Web service attributes(Quality of Service) can be found in the literature references[6-8].The paper utilized a real Quality of Web Service(QWS) data sets provide by EyhabAI-Masri professor, according to the grey system theory(GST) viewpoint and method, performanced the grey relational computation of

nine factors of QWS data sets, and draw the grey relational degree sequence for the optimal selection of Web Service. Figured out the contribution of the factor of QoS. Thus reveals the position and role of the factor of Web Service QoS, based on the practical application of Web services selection provides new ideas and methods. It provides a new thinking and method for selecting Web Service Based on the QoS.

2 Grey System & Grey Relational Analysis

2.1 Grey System Theory(GS)

Grey system theory was pioneered by Deng Julong in 1982, which focuses on “the partial information is known, and the partial information is unknown”, “small sample and poor information” uncertainty system[9]. “the uncomplete and uncertain information” refers to: (1) System factors are not completely cl ear; (2) Factors stoichiometric relation is not completely clear; (3) System structure is not completely clear; (4) System mechanism is not completely clear.

The grey system theory argues that objective system is very complex and characterization of data representation is intricacy, but the system lurks intrinsic rules, and its factors have integral function. Differing form the probability theory and the fuzzy set theory, the grey system theory has obvious characteristic: (1) the small sample and uncertainty, (2) Gray hazy set, (3) information coverage, (4).Many Angle.

2.2 Grey Relational Analysis(GRA)

The grey relational analysis theory is the most mature, the most widely used, and the most dynamic component of grey system theory, which ,in fact, might be reckoned as a contrasting way, in wholeness, equipped with reference for contrasting. It provides a simple cheme to analyze the series relationships or the system behavior, even if the given information is few. It has quantitative analysis and sequence analysis characteristic, also it can make the main factors and secondary factors from the random and unorder sequence, can analysis and determine the degree of the influence factors against the target factors or contributing factor. Its essence is the quantitative comparison analysis of influence factors in the system’s dynamic development trend[10,11].

Procedure of grey relational analysis is as follows:

(1) Collect raw data series, and determine the mother factor X0, which is evaluated as follow:

00X [x (k)],=Where k {1,n}∈…，

determine Sub-factors Xi, which is evaluated as follow:

X [x (k)],i i = Where {1,m},k {1,n}i ∈∈…，…， (2) Determine the spaces of factors @GRF, which is evaluated as follow:

11X (k)/(k)n i i i k x x n ='=∑， Where

{1,m},k {1,n}i ∈∈…，…，

(3) Determine the GR, which means the difference information space in GRA. It can be evaluated as follow:

oi 0(k)|x (k)x (k)|i ?=-，Where {1,m},k {1,n}i ∈∈…，…，

(4) Compute grey relational coefficient , which can be predicated on GR by means of the neighborhood nature and normal interval nature, the algorithm on grey relational coefficient is evaluated as follows:

00(min)*x(max)(x (k),x (k))(k)*x(max)

i i x r ??+=?+ Where {1,m},k {1,n}i ∈∈…，…，

0(max)max max (k)i i k x =?{1,m},k {1,n}i ∈∈…，…，

[0,1],?∈distinguishing coefficient ， usually ，0.5?=

(5) compute grey relational grade , sequence, and weight. By focusing the at utter points the algorithm on grey relational grade is evaluated as follow:

001

1(x ,x )(x (k),x (k))n

i i k r r n ==∑, where {1,m},k {1,n}i ∈∈…，…，

3 Grey Relational Analysis on the Factors of QoS of Web Service

3.1 the Factors of Quality of Service of Web Service

The study utilized a real QWS data sets: QWS Dataset (1.0)[12], which is developed by EyhabAlMasri and Dr. Qusay H. Mahmoud. They have randomly collected 365 Web services from their service repository using Web Service Crawler Engine (WSCE), updated on November,27,2007. This dataset represents 365 real Web services that exist on the Web. The majority of Web services were obtained from public sources on the Web including Universal Description, Discovery, and Integration (UDDI) registries, search engines, and service portals[13-15]. The public dataset consists of 365 Web services each with a set of nine Quality of Web Service (QWS) attributes (We called QWS parameter factors in this paper)that they have measured using commercial benchmark tools. Each service was tested over a ten-minute period for three consecutive days. Each row in the dataset corresponds to an existing Web service implementation available on the public Web today.

QWS(1.0) parameter factors include: Response Time(ab. RS, written as X1): Time taken to send a request and receive a response; Availability(ab. AB, written as X2): Number of successful invocations/total invocations; Throughput(ab. TP, written as X3): Total Number of invocations for a given period of time; Success ability(ab. SU, written as X4): Number of response/number of request messages; Reliability(ab. RE, written as X5): Ratio of the number of error messages to total messages; Compliance(ab. CO, written as X6): The extent to which a WSDL document follows WSDL specification; Best Practices(ab. BP, written as X7): The extent to which a Web service follows WS-I Basic Profile; Latency(ab. LA, written as X8): Time taken for the server to process a given request; Documentation(ab. DO, written as X9): Measure of documentation (i.e. description tags) in WSDL. Partial raw data of QWS are show in Table 1.

Table 1 Partial raw data of QWS

3.2 Grey Relational Analysis Procedure

The polarities of factors in table1 are:X1,X8, Pol(min),X2,X3,X4,X5,X6,X7,X9 Pol(max). Pol(max) means stands for the maximum polarity, which implying that, the more large the samples size, more close to goal, such as Availability, Throughput, Success ability, etc. Pol(min) means stands for the minimum polarity, which implying that, the more small the samples size, more close to goal, such as Response Time, Latency.

In order to provide uniformly effect samples for grey relational analysis, and lay the values in the unit interval (0,1] we should unify the polarities of factors, we thus have algorithms on these factors shown as follow:

min ()1,8:(),{1,}()k

Xi k X X Xi k k Xi k =∈…，365

()2-7,9:(),{1,}max ()k

Xi k X X Xi k k Xi k =∈…，365 Thus the procedure of grey relational analysis is going as follow:

(1) determine the mother factor X0, which is evaluated as:

X0=[X0(k)],Where ,{1,}k ∈…，365

Because all the values is in the unit interval (0,1], We set X0=(1,1,1,1,1,1,1,1,1),which means value 1 is the best for our QoS. The Sub-factors Xi, which is evaluated as follow:

X0=[X0(k)],Where i {1,}{1,}k ∈∈…9，…，365 (2) Determine the spaces of factors @GRF, which is evaluated as follow:

i i 1

1Xi =x ()/x ()n

k k k n ='∑，Where i {1,}{1,}k ∈∈…9，…，365 (3) Determine the GR, which is evaluated as follow:

00i ()|()x ()|i k x k k ?=-，Where i {1,}{1,}k ∈∈…9，…，365 (4) Compute grey relational coefficient , which is evaluated as follows:

0i 000.00180.5*4.2691 2.1364((),x ())()0.5*4.2691() 2.1346

i i r x k k k k +==?+?+ where x(min)=0.0018, x(max)=4.2691,

0.5,?=i {1,}{1,}k ∈∈…9，…，365

(5) compute grey relational grade(GRG), grey relational sequence(GRS), and weight. the algorithm on grey relational grade is evaluated as follow:

0i 0i 1

1(,x )((),x ())n

k r x r x k k n ==∑，where

i {1,}{1,}k ∈∈…9，…，365 then, we can get the final result shown in Table 2

Table 2 Algorithm on Factors of Quality of Service

4 Result Analysis

4.1 Grey Relational Grade & Grey Relational Sequence

Table 2 indicates the grey relational grade (GRG) & grey relational sequence (GRS) of the study, From Table 2 we have r1=0.8004, r2=0.9232, r3=0.7773, r4=0.8954, r5=0.8921, r6=0.9602, r7=0.9727, r8=0.8035, r9=0.7624. Ranking grey relational grades is :r7 r6 r2 r4 r5 r8 r1 r3 r9. According to the GRG, we can classify nine factors into four kinds factors: (1) the first important factors are Best Practices, Compliance, and Availability(with GRG 0.9727,0.9602 and 0.9232 respectively); (2) the second important factors are Success ability, Reliability(with GRG 0.8954

and

0.8921 respectively); (3) the third factors are Latency, Response Time(with GRG 0.8035 and 0.8004 respectively); (4) the fourth factors are Throughput, Documentation (with GRG 0.7773 and 0.7624 respectively).

4.2 Grey Relational Weight

Table 3 shows the grey relational weight (GRW) of the study. From the table 3, we can see the weight sequence (WS) is the same as the grey relational sequence.

Table 3 Grey Relational Weight

In the four kinds of factors above, the first important factors contribute GRW is 36.68% to QoS; the second important factors contribute GRW is 22.96% to QoS; the third important factors contribute GRW is 20.6% to QoS; the less import factors contributes GRW is 19.77% to QoS.

5 Conclusion

The study found the GRG for each factor: Response Time is 0.8004, Availability is 0.9232, Throughput is 0.7773, Success ability is 0.8954, Reliability is 0.8921, Compliance is 0.9602, Best Practices is 0.9727, Latency is 0.8035 Documentation is 0.7624. The sequence of influencing factors of Web Service selection is Best Practices, Compliance, Availability, Success ability, Reliability, Latency, Response Time, Throughput, Documentation.

According to the GRG, the nine factors can be classified into four kinds: the first important factors are Best Practices, Compliance, and Availability; the second important factors are Success ability, Reliability; the third important kind factors are Latency, Response Time; the less important factors are Throughput, Documentation. We also found that the percent contributions of four kinds factors to the QoS of Web Service is 36.68% , 22.96% , 20.6% , 19.77%.

基于Web服务质量因素的灰色关联分析

摘要

本文利用灰色关联分析理论，这一探索性的研究，找到了Web服务质量和其贡献的主要影响因素。本文使用的是从Web服务数据集中所获取的有效数据：QWS数据集（1.0），来进行灰色关联分析。结果表明：（1）影响Web服务选择的因素的顺序是最佳方案，合理性，可用性，成功的可能性，可靠性，延迟，执行时间，吞吐量，资料；（2）这9个因素可以分为四类：其中最重要的因素是最佳方案、合理性和可用性；其次是成功的可能性，可靠性；然后是延迟，执行时间；不太重要的因素是吞吐量，资料；（3）这四类因素对QoS的Web服务质量贡献率分别为：36.68％，22.96％，20.6％，19.77％。

关键词：Web服务；服务质量；灰色理论；灰色关联分析

1简介

Web服务是基于服务体系的结构，它是一种新兴的分布式计算技术[1,2]，并受到了业界和学界越来越多的关注。随着Web服务技术的发展与广泛应用，已经有大量的Web服务技术应用到了因特网中[3-5]，若Web服务搜索仅仅根据功能需求是无法提供用户所期望的服务。Web服务的服务质量与功能属性起着同等重要的作用，因此如何为用户选择相同功能的服务是一项值得研究的课题。有关Web服务的属性（服务质量）可以在参考文献[6-8]中找到。本文利用由EyhabAI 马斯里教授提供的Web服务（QWS）真实数据集，根据灰色系统理论（GST）的观点和方法，求出了QWS数据集中9个因子的关联度大小，并通过灰色关联序为Web服务做出了最优规划。找到了服务质量影响因素的贡献率。，从而揭示了Web服务中服务质量因素的地位和作用，并为Web服务的实际应用提供了新的思路和方法。

1灰色系统和灰色关联分析

2.1灰色系统理论（GS）

灰色系统理论是由邓聚龙于1982年首次提出来的，它主要针对于“一部分信息已知，一部分信息未知”，“小样本，贫信息”等不确定性系统[9]。其中“不完全和不确定的信息”是指：（1）系统因素的不确定; （2）化学计量因素关系

的不确定; （3）系统结构的不确定; （4）系统的机制尚的不确定。

灰色系统理论认为，目标体系是非常复杂的，因此很难用数据来表示其特征，但系统潜藏着的内在规定性，而且这些因素具有整体性。灰色系统理论与概率理论和模糊数理论不同，它具有明显的特点：（1）小样本和不确定性；（2）灰色模糊集；（3）信息覆盖面；（4）多角度。

2.2灰色关联分析（GRA ）

灰色关联分析理论是灰色系统理论中最成熟，应用最广泛，最有活力的组成部分，事实上，它在整体中起到了对比参考的作用。即使给定的信息很少，它也能提供一个简单的方法来分析因素之间的关系或系统的行为特征。它既可以进行定量分析也能进行序列分析，还可以从随机或无序序列中获取主要因素和次要因素，同时也能分析这些因素对主要因素起到阻碍作用或促进作用的程度。其实质是对系统的动态发展过程中的影响因素进行定量分析。[10,11]

灰色关联分析步骤如下：

（1）收集的原始数据序列，并定义主因素序列x 0，分析如下：

00X [x (k)],=其中k {1,n}∈…，

子因素序列X i ，分析如下：

X [x (k)],i i =其中{1,m},k {1,n}i ∈∈…，…，

（2）定义因素序列@ GRF ，分析如下：

11X (k)/(k)n i i i k x x n ='=∑，其中

{1,i ∈∈…，…，

（3）定义GR ?为GRA 中的差异信息空间。它可以作为后续评估：

oi 0(k)|x (k)x (k)|i ?=-，

其中 {1,m },k {1,i ∈∈…

，…， （4）计算灰色关联系数，可通过GR 邻域性和正常的区间性来求得，灰色关联系数的计算如下：

00(min)*x(max)(x (k),x (k))(k)*x(max)i i x r ??+=?+

其中{1,m},k {1,n}i ∈∈…，…，

0(max)max max (k)i i k

x =?{1,m},k {1,n}i ∈∈…，…， [0,1],?∈分辨系数，通常取0.5?=

（5）计算灰色关联度0(x ,x )i r ，序列，和权重。以求0(x (k),x (k))i r 为出发点灰色关联度算法评价如下：

001

1(x ,x )(x (k),x (k))n

i i k r r n ==∑, 其中{1,m},k {1,n}i ∈∈…，…， 2 基于Web 服务的服务质量的因素的灰色关联分析

3.1 Web 服务的服务质量因素

该研究利用一个由EyhabAlMasri 和库塞H.马哈茂德博士开发的QWS 数据集：QWS 数据集（1.0）[12]中的数据。他们从更新于2007年11月27日的WSCE 中随机选取365天的Web 服务数据。大多数Web 服务是从通用，发现和集成注册（UDDI ），搜索引擎和门户网站等公共网络上获取资源。[13-15]公共数据集包括365 个Web 服务集合，每个Web 服务设置九个通过商业基准测量工具确定的质量（QWS ）属性（在本文中QWS 称为参数因素）。我们对每个服务都进行了连续3天的测试，每次测试都进行了10多分钟。现今公共网络中的Web 服务都与数据集一一对应。

QWS （1.0）参数的因素包括：执行时间（ab ，RS ，记为X 1）：从发送请求到接收响应所需要的时间；（2）可用性（ab. AB ，记为X2）：成功调用的数量/总调用数；（3）吞吐量（ab. TP,记为X3）：在给定时间内的总调用数；（4）成功率（ab. SU,记为X4）：响应数/请求信息总数；（5）可靠性（(ab. RE,记为X5）错误信息在所有信息中所占比；（6）合规性（ab. CO,记为X6）：一个WSDL 文档遵循WSDL 规范的程度；（7）最佳实践（ab. BP,记为X7）一个Web 服务遵循WS-1基本要求的程度；（8）延迟（ab. LA,记为X8）：服务器处理一个给定请求的时间；（9）文档（ab. DO,记为X9）WSDL 中的测量文件（例如：描述标签）QWS 的部分原始数据如表1所示。

QWS 表1部分原始数据

3.2灰色关联分析过程

表一中的极值因素：X1,X2(极小值)，X2,X3,X4,X5,,X6,X7,X9(极大值)，max 表示最大极性，即该样品的尺寸越大，越接近目标，如：可用性，吞吐量，成功率等。Min 表示最小的极性，即该样品的尺寸越小，则越接近目标，如：响应时间，延迟。要进行灰色关联分析需要将样本数据统一化，并将关联度大小控制在（0,1]之间，我们对因数序列作如下变换：

min ()1,8:(),{1,}()k

Xi k X X Xi k k Xi k =∈…，365

()2-7,9:(),{1,}max ()k

Xi k X X Xi k k Xi k =∈…，365 因此，灰色关联分析的过程如下：

（1）确定参考序列X0，其评价为：

X0=[X0(k)],其中,{1,}k ∈…，365

因为所有的值是在单位时间间隔（0,1]内，令X0 =（1,1,1,1,1,1,1,1,1），即值1是最适合我们的服务质量。其评估如下：

X0=[X0(k)],其中i {1,}{1,}k ∈∈…9，…，365

（2）确定的因素@GRF ，后续评价为： i i 1

1Xi =x ()/x ()n

k k k n ='∑，其中i {1,}{1,}k ∈∈…9，…，365 （3）确定GR ，其评价如下：

00i ()|()x ()|i k x k k ?=-，其中i {1,}{1,}k ∈∈…9，…，365

（4）计算灰关联系数，其计算如下：

0i 000.00180.5*4.2691 2.1364((),x ())()0.5*4.2691() 2.1346

i i r x k k k k +==?+?+

其中，X(min)=0.0018，X(max)=4.2691，

0.5,?=i {1,}{1,}k ∈∈…9，…，365

（5）计算灰色关联度（GRG ），灰色关联序列（GRS ），和权重，灰色关联度的算法计算如下：

0i 0i 1

1(,x )((),x ())n

k r x r x k k n ==∑， 其中i {1,}{1,}k ∈∈…9，…，365

然后，我们可以得到表2中所示的最终结果

表2服务质量的因素

4 结果分析

4.1 灰色关联度及灰色关联序

表2显示了灰色关联度（GRG ）与灰色关联序列（GRS ）之间的关系，从表2中，我们有，r1=0.8004, r2=0.9232, r3=0.7773, r4=0.8954, r5=0.8921, r6=0.9602, r7=0.9727,

r8=0.8035, r9=0.7624.灰色关联序为：:r7> r6 >r2 >r4> r5> r8 >r1 >r3> r9.根据GRG 值我们可以将这9个因子划分为4类因素：（1）最重要因素是最佳实践，合规性和可用性（GRG 分别为0.9727,0.9602，0.9232）；（2）第二重要因素是成功率，可靠性（GRG 分别为0.8954和0.8921）；（3）第三重要因素是延迟，响应时间（GRG 分别为0.8035和0.8004）;（4）第四重要因素是吞吐量，文档（

GRG

分别为0.7773和0.7624）。

4. 2 灰色关联重量

表3给出了研究的灰关联权重（GRW）。从表3中我们可以看到权重序列（WS）和灰色关联序列相似。

表3灰关联权重

在以上四种因素中，对QoS贡献度最大的因素占33.68%，第二重要的因素占22.96%，第三重要的因素占20.6%，最不重要的因素占19.77%。

5 结论

这项研究发现，GRG的每个因素中：执行时间是0.8004，可用性是0.9232，吞吐量是0.7773，成功率为0.8954，可靠性是0.8921，合规是0.9602，最佳实践是0.9727，延迟是0.8035，文档是0.7624。因此选择影响因素的顺序是最佳实践，合规性，可用性，成功率，可靠性，延迟，响应时间，吞吐量，文档。

根据GRG的分析结果可知，九个因素可以分为四种：第一个重要的因素是最佳实践，合规性和可用性;第二个重要因素是成功率，可靠性;第三类重要的因素是延迟，响应时间;不太重要的因素是吞吐量，文档。我们还发现，四种因素对网站的服务质量影响百分比是36.68％，22.96％，20.6％，19.77％。

本文来源：https://www.bwwdw.com/article/4y8l.html