Multicriteria Decision Making Taxonomy of Cloud-Based Global Software Development Motivators

The software organizations widely consider the cloud based global software development (CGSD) as it offer the quality projects with low cast. The adoption of CGSD is challenging due to the geographical distance between practitioners. This study aims to identify and analyses the motivators that could positively impact the implementation of CGSD paradigm. The systematic literature review approach was applied to identify the CGSD motivators reported in the literature, and were further validated with industry experts using questionnaire survey study. Moreover, the fuzzy-AHP approach was applied to prioritize the investigated motivators concerning their significance for the successful adoption of CGSD. The findings of the study provide the prioritization-based taxonomy of the investigated motivators that assists the software organizations to develop and revise their strategies for the successful implementation of CGSD.


I. INTRODUCTION
The cloud computing is increasingly adopted in the geographically distributed software development environment as it provides significant opportunities to execute and manage the software development process. The availability, scalability and the dynamic attracted th e software firms to consider the cloud based global software development (CGSD). Dhar [1]stated that in software industry the outsources includes the development practices, process and decision management and the services were transformed in different geographical location across the globe.
Currently, the adoption of CGSD paradigm is significantly increased [1]. Fan et al. [2] mention that the CGSD paradigm educate the software development organization in terms of marked demand and the future trends.
The associate editor coordinating the review of this manuscript and approving it for publication was Yang Liu . Clemons and Chen [3] argued that it is necessary to take the rite decision and rite time for development of quality projects. They also mentions that CGSD assists to make the right decisions considering the trend and demand of international market. Chang and Gurbaxani [4] mention that it is important to make the right decision at right time contributed to develop the quality projects within time and budget. In this study, the definition of Leimeister et al. [5] is used i.e. ''an IT deployment model based on virtualization, where resources in terms of infrastructure, applications and data are deployed via the internet as a distributed service by one or several service providers. Services are scalable on-demand and can be priced on a pay-per-use basis.'' The development of good quality projects with low cost and time is always the priority of every software development organization. Though, the CGSD provides the opportunity to achieve to software organization to achieve such objective by hiring the skilled human resource from developing countries and by arranging the development activities round the globe [6], [7]. The services of cloud computing assist the software organization to outsource their development activities by providing the advanced tools and technologies [7].
Thus, the adoption of CGSD paradigm is not straightforward and various complexities are faced by the software organizations concerning to the successful execution of CGSD paradigm.
Niazi et al. [8] indicted that the geographically distributed teams experienced the different challenges compared with collocated development environment. For example, the communication and coordination both are the important activities of CGSD paradigm, though, the physical distance between the practitioners hindering the effective communication and coordination [9].
Liu and Wang [10] indicated the language, culture and the temporal distance between the CGSD teams is also one the key problems of poor communication and coordination. They further point that the limited physical meetings communication and meeting also cause the lack of trust between the overseas practitioners. Various other studies also highlighted the hindering factors of CGSD paradigm such as: time zoon differences, lack of process synchronization, delay in response etc. [11], [12].
Besides the importance of CGSD, little attention has been given to empirically identify the motivators that could positively impact the CGSD activities. Hence, the aim of this study is to explore and analyze the motivators of CGSD. Therefore, the study objective has been address by applying the following steps: (i) to explore the CGSD motivators from the literature via systematic literature review (SLR), (ii) validate the identified motivators with industry experts using a questionnaire survey approach and (iii) analyses the identified motivators using the fuzzy-AHP approach and develop the prioritization-based taxonomy of the CGSD motivators. We believe that the deep understanding and analyses of the CGSD motivators assists the academic researcher's real-world practitioners to develop the new strategies and techniques for the success and progression of CGSD paradigm. This study addresses the following research questions: RQ1: What are the important motivators of CGSD as reported in the literature? RQ2: What practitioners do think about the CGSD motivators identified via literature review?
RQ3: How to prioritize the investigated motivators? RQ4: What would be the prioritization based-taxonomy of the investigated motivators?
The rest of the paper is structured as: the related work is discussed in section 2. Section 3 contains the adopted research methodologies. Results and analysis are presented in section 4. Summary and discussion are given in section 5, and section 6 contains the thread to validity. Future work and conclusion are summarized in section 7.

II. RELATED WORK
The cloud computing provides the services concerning to the requirements of clients. The cloud is structured to provides the efficient and easy access the well managed resources, organized by services providers [13], [14].
The distributed nature of cloud services offer an opportunity to adopt global software development phenomenon [14].
The CGSD offers the availability of skilled labor and round the clock development times and these attributed motivated the software organizations to adopted geographically distributed development environment [8].
Kahraman et al. [15] mention that in developing countries, the development wages are one third low compared with developed countries. Similarly, Heininger [16] underlined that due to the economic benefits, the client organizations of developed countries outsourcing their development activities to developing countries. Hence, the cloud computing make the outsourcing paradigm more easy and reliable as it provides the virtual accessibility of data and resources across globe [17]. The CGSD paradigm assists to educate the software practitioners with the updated tools and technologies used in developed countries and the market trends [16], [18].
Besides, Jugdev et al. [19] mention that the adoption of CGSD is beneficial and challenging at the same time. As in CGSD environment, the practitioners are involved across the globe having different cultural ethics, language and time zones, which causes the lack of effective information sharing [20]- [22]. Yang et al. [23] also mention that in CGSD environment, the communication and coordination are badly effect due to the lack of frequent communication and coordination.
Considering the existing literature, no research has been conducted to explore the motivator of CGSD paradigm. Hence, there is a dire to explore the motivator that could positively impact the CGSD paradigm. Moreover, using the fuzzy-AHP, the identified best practices were prioritized and develop the prioritization based taxonomy [24]. The fuzzy-AHP approach has already been applied in by several research domain to address the complex decision-making problems [25]- [29], etc. In the current study, we have also applied the fuzzy-AHP approach to priorities the motivators of CGSD based on the experts opinions. The prioritization also provides the robust taxonomy of the investigated motivators that could help the practitioner to consider the highest-ranked motivators for the successful execution of CGSD process.

III. RESEARCH METHODOLOGY
To address the study goals, the research was designed in three different steps. In the first step, the systematic literature review approach was applied to explore the important motivators reported by the researchers in the existing literature. Secondly, the findings of the literature review were empirically validated with industry experts using a questionnaire survey approach. In the final step, the fuzzy-AHP approach was applied to prioritize the investigated motivators concerning to their significance for the success and progression of CGSD. The adopted research design is given in Figure 1 and are explained in the following sections.

A. SYSTEMATIC LITERATURE REVIEW (SLR)
To explore the potential literature aiming to address the research objective of this study, the step by step protocols of SLR ware adopted [30]. An SLR is a well-established approach to collect the most related studies related to the study RQs. The step by step protocols of the SLR approach assists in identifying and evaluating the primary studies related to the specific research area [31]. In this study, we adopted the SLR protocols developed by Kitchenham and Charters [31]. The findings of SLR are more valid and comprehensive compared with informal literature reviews. The SLR is a widely adopted research methodology in software engineering [8], [32], [33]. The phase of the SLR approach is given in Figure 1 and discussed in the following sections:

1) RESEARCH QUESTIONS
This study aims to explore the motivators of CGSD from the existing literature. The research question (RQ1) of this study are presented in section-1.

2) SEARCH SOURCES
To collect the most appropriate and potential literature, the selection of appropriate data sources are important [34]. By following the instructions of Chen et al. [34] and Niazi et al. [8], the following database were selected for the search process. The selected database covers the major areas of mainstream research publications.

3) SEARCH STRING
To explore the most related literature from the above enlisted database, the development of appropriate search string is important. Though, the following search string was develop using the keywords and their alternatives, aiming to explore the most potential literature from the selected databases. The keywords and their alternative were collected from the existing papers, i.e. [4], [8], [14], [16], [17], [35]. To formulate the search string, the Boolean ''OR'' and ''AND'' were used as presented below: Infrastructure as a Service'' OR ''Platform as a Service'' OR ''Software as a Service'' OR ''IT service'' OR ''Application Service'' OR ''ASP'').

4) INCLUSION AND EXCLUSION CRITERIA
For the further refinement of collected, we have developed the inclusion and exclusion protocols. These protocols were developed by considering the other SLR studies [8], [36].
For the inclusion: we only consider the studies published in the mainstream journal and conferences; the selected studies should be a book chapter, conference, or journal paper; the article should be in English; the articles based on primary investigations.
For exclusion: the studies that particularly not focus on CGSD; the article does not have a detail discussion about the reported CGSD motivators, and the studies were also not considered who have duplication in results; moreover, if the studies are from same research project, the final one is consider in our SLR.

5) STUDY QUALITY ASSESSMENT
The selected literature was assessed aiming to determine the significance of final collected literature with respected to the study RQs. We have formulated the checklist of study quality assessment using guidelines of Kitchenham and Charters [31]. The developed checklist was consisting of five questions, as presented in Table 1, and every question was evaluated via Likert scale given in Table 1. All the selected studies, and their quality assessment score, are provided in Appendix-A.

6) STUDY SELECTION
Furthermore, the tollgate approach developed by Afzal et al. [37] has been adopted aiming to the final refinement of selected materials, for considering in data extraction process. Initially, 1239 studies were collected in-response of the execution of search string on the selected databases, after applying the inclusion and exclusion criteria (given in section 3.1.4). Therefore, we carefully performed the steps of the tollgate approach (Figure 2), and finally,67 studies were selected for the final data extraction process. All the studies were labelled with SL and are given in appendix-A.

7) DATA EXTRACTION AND SYNTHESIS
The final selected 67 studies ware used for data extraction process. The data were extracted while performing the steps of tollgate approach. In the collection process, the first and third author of this study were continues involved. Second and fourth author of this study verify the data extracted data and perform all the steps of the tollgate approach. Initially, all the statements, ideas, practices, factors were extracted from each paper. The extracted data were carefully reviewed to address the duplication and rephrase them to avoid similarity concern.
In addition, the inter-rater reliability test Afzal et al. [37] was conducted with the aim to determine the researchers biasness. To do this, three experts of other software engineering domain were invited and ask them to perform the data extraction process. They randomly selected 15 papers and performed all the steps of tollgate approach and data extraction. Based on the result of study researchers and the external participants, the non-parametric Kendall's coefficient of concordance (W) [37] was determined. Whereas, W = 1 shows that complete agreement and W = 0 presented the no agreement. Thus, the determined (W = 0.84 (p = 0.003) presents the significant agreement between the findings of both teams. Thus, the findings of SLR study is consistent.

8) REPORTING THE REVIEW a: QUALITY ASSESSMENT QA) OF SLR STUDIES
To check the appropriateness of considered literature, the QA was performed considering the developed criteria ( Table 1). The results of QA shows that 83% of the selected studies scored grater 65%. This indicated that the selected literature has the potential to effectively address the research questions and objectives [33], [38]. The detailed results of QA process are given in Appendix-A.

b: ADOPTED RESEARCH METHODS IN THE SELECTED STUDIES
The research techniques applied by the selected literature were also determined during data extraction process. The purpose of research approaches extraction is to check whether the findings of selected studies based on empirical data or not. Though, It is noted that form the 67 selected studies: 36% studies used questionnaire survey (QS), 28 used case study (CS) method, grounded theory (GT) and content analysis (CA) considered by 6% and 9% studies, respectively, action research (AR) approach was adopted by 8% of the selected studies and mixed-method (MM) used by 13% Figure 3). According to the percentage analysis, majority of the selected studies were used the QS approach.

B. EMPIRICAL STUDY
As the CGSD is purely practical oriented software development paradigm, though, it is important to verify the findings of SLR study with industry practitioners. The steps adopted to verify the SLR findings are presented in the sub-sequent sections.

1) DEVELOPMENT OF SURVEY INSTRUMENT
The survey instrument was developed to get the opinions and perceptions of industry experts regarding the identified motivator from SLR study. The survey instrument was developed using the Google Form platform. An online questionnaire survey is an effective approach to collect the opinions form geographically distributed practitioners around the world [36], [39]. The survey instrument was divided into three steps. The first section consists of the bibliographic information of survey participants. The second section contain the questions related to the identified list of motivators. The development questionnaire also contain an open-ended section, which allows the survey participants to put the additional motivators that are not enlisted in the second section. The five point Liket scale ''strongly agree,'' ''agree,'' ''neutral,'' ''disagree,'' and ''strongly disagree,'' was used to get the opinions of survey participants. It is important to provide a neutral option in the Liket scale, as instead of this, the respondents are bound to make the decision one sided.

2) DATA SOURCES
The aim of this empirical study is to verify the list of motivators identified via SLR study. The selection of potential population is important to get the most effective data from the participants. As the basic domain of this paper is global software development, thus, we approach the participant working in such domain and working across the globe. Hence, the ''exponential discrimination snowball sampling'' technique [41] has been followed to target the most related population according to the study domain. The snowball sampling is an effective and easy way to collect the data from most is an easy and cost-effective way to collect data from the geographically distributed population [42].
Watters and Biernacki [43] also argued that the snowballing is an effective technique to collect the data from dispersed population.
We have used the Email, LinkedIn, Research-Gate as a sources for data collection [8], [36]. the data were collected during December-2019 to March-2020. During data collection process, 97 response were collected. The collected response were checked to found incomplete entries and 8 response were found uncompleted.
The complete 89 response were considered for further analysis. Appendix-C presents the detailed bibliographic information of survey participants.

3) SURVEY DATA ANALYSIS
The frequency analysis approach was used to analyze the feedback received from the survey experts, as it is the effective method to compare the variable within and across the categories [44]. Various researchers of empirical software engineering used the same approach [8], [36], [45].

4) FUZZY AHP SURVEY
This study aims to identify the motivators of CGSD paradigm. Though, we have applied the fuzzy-AHP technique to prioritize the identified list of motivators with concerning to their criticality for CGSD paradigm. To develop the pairwise comparison matrixes, we have conducted a fuzzy-AHP survey. To do this, we have contacted the participants of first survey, and send them pairwise questionnaire survey (Appendix-D). In response, we have collected 26 complete responses and that were used for further data analysis process. The collected 26 responces might be reflected as small data sample. As the fuzzy-AHP method is a subjective approach and it also accept the small data sample [26], [46]- [48].
Using the geometric mean, the data collected data were converted into TFN number which are used to develop the pairwise comparison matrixes. To determine the geometric mean, we have considered the following formula: The basic fuzzy set theory and procedure of AHP approach are discussed in this section. To address the MCDM concern, the AHP, Fuzzy-AHP and the Fuzzy-TOPSIS are exist. In this study, the fuzzy-AHP approach was considered as it is more efficient approach to address the MCDM problems [50], [51]. The simple AHP is not effective enough to address the ambiguity and vagueness of experts. Though, the combination of fuzzy-AHP is the best technique to critical analyses and address the MCDM problems [52], [53].

1) FUZZY SET THEORY
Zadeh [54] developed a fuzzy set theory which is an extended version of conventional set theory. This approach is oriented to address the vagueness and uncertainties faced by the real-world industry practitioners while making MCDM. The Zadeh approach effectively work to deal with the uncertainties and vagueness [55,56]. In this approach, a characteristic function µ V (x) is embedded, which assists to map the membership of provided value range from 0 to 1. The concept and introduction of the fuzzy set are presented below: Definition: Triangular fuzzy number (TFN) V is renders by a triplet (vl, vm, vu). The characteristic function µV(x) of a TFN is shown in equation (1) and Figure 4.
where v l indicates the lowest,v m shows the most favourable, and v u present the highest ranked value. The most commonly adopted algebraic functions in to TFN operations s, say (V 1 , V 2 ) are given in Table 2.

2) FUZZY ANALYTICAL HIERARCHY PROCESS (FUZZY AHP)
To deal with MCDM problems, fuzzy-AHP is consisted as the most appropriate approach. The features of fuzzy-AHP has the ability to fix both qualitative and quantitative data. The following steps were adopted to perform the fuzzy-AHP: Phase1: Develop the hierarchy structure of the problem ( Figure 4). Phase2: Develop the pair-wise comparison matrixes to determine the priority vector.
Phase3: Perform the consistency check by determining the consistency ratio.
Phase4: Determination of priority rank of categories and the sub-categories (i.e. motivators) ( Figure 5).
Although, the conventional AHP has various merits, but its usage is narrow in the situations where the generated data are uncertain and vague. Hence, in such conditions, the fuzzy AHP is the most appropriate methodology and its results are more correct [55].The same approach has been applied by various existing studies to fix the uncertainties and vagueness [26], [46]- [48]. In this study, we have applied theChang's method [57] because of its effectiveness and acceptance in research community. Chang [57]expressed a ranking problem as collection of objects which are called main categories and represented as X = {x 1 , x 2 , . . . , x m }. Each x i ''also contains elements, called goal set, and represented as V = {v 1 , v 2 ,.., v n }. At a time, one main category, x i, is considered, and each goal g i undergoes extent analysis. Therefore, each category undergoes extent analysis (m) times which can be calculated'' using the following Equation (2) and (3) [57]: Phase 1: The analysis of fuzzy extent of the i th category is presented in Eq. (4) as: can be determined as presented in equation (6) and (7): VOLUME 8, 2020 Phase 2: Given two TFNs V a and V b , degree of possibility that V a ≥ V b can be defined as: Or more specifically as: where d presented the highest value of intersection in µ Ga and µ Vb (Figure 6). Phase 3: The overall degree of possibility of a given convex fuzzy number H is determined concerning other V i (i = 1, 2,. . . , k) as: Assuming that, where k = 1,2 ,. . . ,n and k = i.
Eq. 11 is used to determine the weight vector as: where, V i (i = 1, 2, . . . , n) are n separate fuzzy numbers. Phase 4: The weight vector W determined using eq.12 is normalized to achieve priority weight as a crisp number: here, W represents a crisp number. Phase 5: Checking consistency ratio: the consistency is a necessary aspect of all the pairwise comparison in the fuzzy AHP analysis method [27]. Therefore, we have conducted the consistency check for all the pair-wise comparison matrixes. To do this, all the matrixes were transformed into corresponding crisp values by applying the graded mean integration method and this is called defuzzification. Afterwards, we have used the following formula to transform a TFN P = (l, m, u) to an equivalent crisp number: Once P crisp is calculated, the consistency index (CI) and the consistency ratio (CR) is determine as: where, I max : the maximum eigen-value of the given comparison matrix, n: number of element in a matrix. RI: Random index value (Table 3). If the determined CR >0.1, the pairwise matrix is content, and the inconsistency require to gain collect the data from experts.

IV. RESULTS AND ANALYSIS
The results and analysis of this study are presented in the section.

A. FINDINGS OF SLR STUDY
By carefully reviewing studies selected using SLR, a total of 39 motivators were extracted. The list of investigated motivators is given in Table 4. The investigated motivators present the critical aspects of CGSD paradigm that need to be focused by the organizations. The identified motivators were further mapped in the project management body of knowledge (PMBOK) areas core areas [58]. All the knowledge area of PMBOK identified the key zones that a project manager should adopt for the successful execution or development activities. All the knowledge areas of PMBOK are classified into three main categories [58]: • C1 (project objective knowledge area): quality, scope, time and cost.
• C2 (knowledge area that facilitates to accomplish the project goals): procurement management, stakeholder management, risk, human resources and communication.
• C3 (knowledge area effected due to other knowledge areas): project integration management. All the knowledge areas are important to the successful execution of project management activities [8]. However, the investigated motivators of CGSD were mapped into ten knowledge area of PMBOK, to provide the body of knowledge to practitioners which assist in developing the strategies for the successful execution of project management activities in the context of CGSD. This classification is also helpful for the researcher to conduct their future research to the most priority areas of project management in the domain of CGSD. The research work is significant to develop the tools and techniques to address the motivators of CGSD process.
The ''coding scheme of Grounded Theory'' [59] technique was adopted to map the explored motivators of CGSD into ten knowledge areas of project management. The mapping team consist of three members; they labelled and grouped the motivators into ten most related knowledge areas. The mapping result (Figure 7) indicated that Human resources management is the most critical knowledge area of the identified CGSD motivators. The practitioners pay more concentration to address the motivators of human resource management category.
To avoid the researcher's bias, we have performed an interrater reliability test between mapping team and indented experts. In this test, two independent experts were involved, and they map all the 39 motivators to ten knowledge areas according to their understanding. We have determined ''non-parametric Kendall's coefficient of concordance (W),'' [60] to determine the inter-rater agreement among the researchers and external experts. The value of W = 1, indicate the complete agreement and W = 0 indicate the complete disagreement. Though the analyzed results (W = 0.86, p = 0.004) indicated that there are significant similarities between the mapping process of researchers and independent experts, this renders that the mapping process is in an agreement between the researcher and independent experts. Hence the categorization is unbiased.

B. RESULTS OF EMPIRICAL STUDY
After the identification and mapping of investigated CGSD motivators, the empirical study was done to get the insight of industry expert. The collected results were broadly classified into three categories, i.e. ''positive, negative and neutral.'' The positive (''strongly agree and agree'') category consists of the percentage of survey respondents who agree with the identified motivators and their categories as they are related to the CGSD paradigm. The negative (''strongly disagree, disagree'') category presents the results of survey participants who did not consider the explored motivation for CGSD paradigm. The result presented in the neutral category shows the responses of those participants, who don't know the impact of an identified motivator or their categorizations.
According to the summarized results of the questionnaire survey study (Table 5), the majority of the survey participants agrees with the identified motivators of CGSD. The results show that M4 (Standardization with the internal process, 90%) is acknowledged as the most important motivator for the success and progression of CGSD organizations. We further noted that M19 (Trust building, 85%) and M39 (Client and vendor interaction, 85%) are considered as the second most important motivators for CGSD paradigm.
The results show that C4 (Scope, 91%) was considered as the most important category of the investigated motivators. This renders that according to the survey participants, the motivators of scope category more significant for the success and progression of CGSD. Moreover, C2 (Time, 90%) and C9 (Procurement, 89%) were considered as the second and third most important knowledge areas of PMBOK for the successful execution of CGSD practices.

C. APPLICATION OF FUZZY ANALYTIC HIERARCHY PROCESS
This section contains the results determined by applying fuzzy-AHP techniques. The calculated priority ranks of the motivator and their core categories are presented in the blow sections.
Step-1: (development of hierarchy structure) In this step, a critical problem is divided into the interconnected decision making element considering the knowledge areas of PMBOK. In the hierarchy structure, the problem is structured into three levels as presented in Figure 5. In the developed hierarchy, the main VOLUME 8, 2020 objective is presented on top level. The categories and their respective sub-categories are presented on level 2 and 3, respectively. The proposed hierarchy structure is given in Figure 8.
Step-2: Pairwise comparison The pair-wise comparison matrixes were developed using the response collected in fuzzy-AHP survey. We further analysed the developed pairwise comparison matrixes to determine the priority weights of each motivator and their respective core categories. Table 6 shows the linguistic variable with respect to their triangular fuzzy Likert scales. To develop the pairwise comparison matrixes of the identified motivators and their categories; the triangular fuzzy conversion scale proposed by Bozbura et al. [61] was adopted. The developed pairwise matrixes of the motivator of each category are presented   in Table 7 , 11, 12, 13, 14, 15, 16, 17, 18, 19 and in between the categories are presented in Table 20.

Step-3: A numerical example for the determination of local priority weight of 'Integration' category motivator
The priority of all the motivator of integration category were determined as presented in Table 7. To determine the local priority weigh (LPW) for all the motivators and their respective categories the Eq.3 was applied. Firstly, we have determined the synthetic extent values for all the motivator fo integration category. Secondly, using the Eq.4, the priority vector was calculated. The calculation of priority weigh of each motivator is presented below.
By Eq.6, the degree of possibility was determined and Eq. 8 were further employed to determine the minimum possibility (LPW) using the pairwise comparison matrixes.
Hence, we have calculated the weight vector as: W = (1, 0.030029, 0.69846, 0.36305) as presented in Table 8. After normalizing these values, the determined importance attributes are W = (0.4789, 0.01435, 0.3337). According to the determined result M1 (Integration with organizational IT infrastructure) is the highest priority vector as compare to the determined weights of other motivators.
Step-4: consistency check This section presents the steps to determine the consistency of pairwise comparison matrixes. To do this, the crisp, fuzzy values of the motivators of 'Integration' category presented in Table 9 are considered. Using the Equation 14, a triangular fuzzy number of the pair-wise comparison matrix of the main categories are defuzzified to crisp number, and get the Fuzzy Crisp Matrix (FCM) as presented in Table 9.'' To determine the value of largest Eigenvector (λ max ), the column sum for all the attribute of FCM matrix was determined (Table 9). Furthermore, each element of FCM matrix divided by column sum. Finally, the average of each row was determined to calculate the final priority weigh as presented in Table 10.
Considering the calculation of largest Eigen-value (λ max ) of the matrix FCM is 4.1067. The dimension of FCM is 4. Therefore n = 4 and the Random Consistency Index (RI) is 0.9 for n = 4 (Table 3). We used the equation 15 and 16 are used to calculate the consistency index and consistency ration     as follows: The calculated value of CR is 0.039503<0.10; which renders that the pairwise comparison matrix of the motivators of integration category is consistent. Using the same procedure, the consistency of all the other pairwise matrixes of were determined and the results are presented in Table 11, 12,  13,14,15,16,17,18,19, and 20, respectively.

Step-5: Local Weight and Global weights calculation
Local weights (LW) of the motivators were calculated considering the weight vectors value (W) in the respective categories (Table 21). The LW used to determine the local ranking of the motivators in their categories.
In 'Integration' category, M1 (Integration with organizational IT infrastructure, LW = 0.379381) is declared as the highest-ranked motivator, because its weight is higher compared with other motivator of 'Integration' category (Table 21). Moreover, M3 (Solid help desk and support structure for overseas sites, LW = 0.275932) and M4 (Standardization with internal process LW = 0.195243) are acknowledged as the third and fourth most critical motivators VOLUME 8, 2020    in 'Integration' category. The local ranking (LR) indicates the priority level or significance of a particular motivator within their respective category. The LR help to deal with motivators of specific project areas. Furthermore, the global weigh (GW) of all the motivator was calculated to evaluate the impact of each motivator on overall CGSD paradigm. All the motivators were globally ranked based on their GW. The GW was determined with the multiplication of local weigh of each motivator and the weight of their respective category. For example, the global weight of M1 = local weight of M1 * category weight (i.e. Integration) = 0.379381× 0.413492 = 0.15687.
Based on the determined GW value of M1, it is ranked as the 1 st most significant motivator for CGSD paradigm (Table 21).
Using the same concept, we have determined the GW for all the motivators and the results are given in Table 21. The GW result shows that M12 (Scalability, GW = 0.16), is noted as the 2 nd ranked motivator for CGSD. Furthermore, M14 (Maintenance and updating, GW = 0.11944) and SF10 (Solid help desk and support structure for overseas sites, GW = 0.275932) are standout the 3 rd and 4 rd are highest ranked motivators for the CGSD environment.

V. SUMMARY AND DISCUSSION
The key objective of this study is to explore and analyses the CGSD motivators with respect to their criticality for CGSD paradigm. To address the study aims, the study has been conducted in three different steps. The results and analysis of each research question are discussed in the following sections: RQ1: What are the important motivators of CGSD as reported in the literature?
By conducting the systematic literature review, a total of 67 studies based on empirical evidence were collected. The final selected studies were carefully reviewed, and 39 motivators were identified. The identified motivators present the key area that needs to be fix for the success and progression of CGSD. The identified list of motivators was further mapped into ten core PMBOK areas. The basic purpose of this classification is to develop a hierarchy structure which assists in implementing the steps of fuzzy-AHP. Moreover, the categorization process also helpful for academic researchers and practitioner to consider the important area of CGSD motivators that need to be addressed for the success and progression of CGSD paradigm.

RQ2: What practitioners do think about the CGSD motivators identified via a literature review?
The identified motivators via literature review were scaled to industry practices using questionnaire survey study. Based on the summarized results of survey study, all the investigated motivators are related to industry practices to some extent. Besides, we also intended to check the validity of the mapping process. The analyzed results of the empirical study show that the mapping is valid and consistent.

RQ3: How the investigated motivators are prioritized?
The fuzzy-AHP was applied to prioritize the investigated motivators and their core categories concerning their significance for CGSD. The fuzzy-AHP is an effective technique to address the vagueness and uncertainties in the expert's opinions. To address this objective, the pairwise comparison among the motivators and their relative core categories. The pairwise comparison is useful to calculate the priorities of identified motivators concerning to their criticality for CGSD. All the investigated motivators and their corresponding categories were ranked considering the calculated weights using fuzzy-AHP. The fuzzy-AHP technique provides a complete understanding of the MCDM problems that incorporate the significance of CGSD paradigm motivators and their respective core categories.
The Fuzzy-AHP results show that M1 (Integration with organizational IT infrastructure) is ranked as the highest-ranked motivator for CGSD. To effectively use cloud services, the compatibility of cloud technology and the internal organizational infrastructure is significant. Hashmi et al. [62] also indicated that the global software development organization should adopt up-to-date development infrastructure for the better integration of cloud services with local IT devices. According to the results, M12 (Scalability) is declared as the 2nd most important motivator for CGSD environment.
Scalability is a cloud service that allow to grow and diminish the resources required to address the business need and capably. In essence, scalability is a planned level of capacity that can grow or shrink as needed. Though, cloud computing provides a flexible environment for a global software development organization to use the services of the cloud concerning their need. Adjepon-Yamoah and David et al [63] and Al-qadhi et al. [64] also indicated the Scalability as the key motivator of CGSD organizations.
Moreover, M14 (Maintenance and updating, GW = 0.11944) and M10 (Solid help desk and support structure for overseas sites, GW = 0.275932) are ranked as the 3nd and 4rd top ranked motivators for the CGSD projects. Oza et al. [65] indicated that the regular maintenance and updating is important for a CGSD site to ensure to use the full capacity of cloud services. Smirnova [66] further mention that timely updating and maintenance of IT infrastructure keep it up-to-date, and that assist to employ the cloud services better.

RQ4: What would be the prioritization basedtaxonomy of the investigated motivators?
Prioritization-based taxonomy of the investigated motivators was developed by using the local and global ranks determined using fuzzy-AHP analysis. The developed taxonomy ( Figure 9) presents the ranks of each identified motivator with comparison to the local and global weights that indicated the significance of a particular motivator within their respective category and for the overall objective of this study.
The developed taxonomy indicated that M1 (Integration with organizational IT infrastructure) was ranked as the highest-ranked motivators with respect to both local and global ranking. This indicated that the organizations should consider the M1 on a priority basis for the successful execution of CGSD paradigm. Similarly, it is noticed that M3 (Solid help desk and support structure for overseas sites) is locally ranked as 2, and globally ranked as 4. An important observation is that the M2 (Improve automation) is locally ranked as 4th and for global context, it is ranked as 25th. However, this variation assist the experts and researchers to consider the most priority motivator concerning their interested category and also by considering the overall study objective. The results show that 'Scope' is the highly ranked category of the identified motivators.
Moreover, 'Integration' and 'Communication' are ranked as the 2 nd and 3 rd most priority motivators categories, respectively. The developed taxonomy indicates the complete picture of the investigated motivators and their criticality in CGSD domain. Study results provides a conceptual framework that could assist the practitioners to scale their software development activities in CGSD environment.

VI. THREATS TO VALIDITY
The data was extracted from the limited digital repositories and this might causes the missing of some related studies. Based on the other studies, this is not a systematic problem [36], [45], [67], [68].
Similarly, the extracted data from the selected studies might be not consistent and have uncertainties. We address this threat by conducting the ''inter-rater reliability test'' and the results shows that there is no researcher's baseness and the extracted data is consistent.
Another potential threat towards the validity of study findings is the majority of the survey participants are form developing countries. Though, we also noted a representative set of survey respondents from developed countries that enables to generalize the study findings.

VII. CONCLUSION AND FUTURE DIRECTIONS
The software development organizations are increasingly transforming their businesses from collocated to geographically distributed development environment using cloud computing services. The increasing interest of software organization in the adoption of geographically distributed software development paradigm motivated us to explore the important motivators of CGSD. Using the systematic literature review approach a total of 39 motivators were identified from the literature. The investigated motivators were further mapped into ten key knowledge areas of the project management body of knowledge (PMBOK). The identified list of motivator and their categories were further verified with industry experts using questionnaire survey approach. The results of questionnaire survey study show that all the identified motivator and their categories are related to the real-world industry practitioners. In the final phase, we have applied the fuzzy-AHP to prioritize the investigated motivator concerning their significance for CGSD. The results of the fuzzy-AHP analysis show that 'integration with organizational IT infrastructure', 'scalability', 'maintenance and updating' and 'solid help desk and support structure for VOLUME 8, 2020 overseas sites' are the highest-ranked motivators that need to be considered on priority for the successful execution of CGSD process. Moreover, the categorization of investigated motivators into ten knowledge areas of PMBOK and the calculated local and global rankings provides a robust taxonomy. The developed taxonomy assists to determine the significance of each investigated motivator within their respective category and for overall study objective. Ultimately, the findings of this study will assist the researcher and practitioners in revising and to the development of new techniques for the successful adoption of CGSD environment.
The future of this study is to investigate the influencing factors (positively and negatively) of CGSD paradigm. Ultimately, a robust readiness model will be developed to assist the CGSD organizations in assessing and improving their development activities.

APPENDIXES APPENDIX-A
List of selected studies and their quality assessment score (https://tinyurl.com/y9j3qgxn) He is also an Active Researcher in the field of software engineering and has published more than 50 articles in peer-reviewed journals and international conferences. He worked as a principal and a co-investigator in a number of research projects that investigate issues related to component based software development and global software development projects. His research interests include empirical software engineering, evidence based software engineering, component-based systems, global software development, and software process improvement in general. ABDU GUMAEI received the Ph.D. degree in computer science from King Saud University, Riyadh, Saudi Arabia, in 2019. He worked as a Lecturer and taught many courses, such as programming languages at the Department of Computer Science, Taiz University, Yemen. He is currently an Assistant Professor with the College of Computer and Information Sciences, King Saud University. He has authored and coauthored more than 30 journal and conference papers in well-reputed international journals. He received a patent from the United States Patent and Trademark Office (USPTO), in 2013. His research interests include software engineering, image processing, computer vision, machine learning, networks, and the Internet of Things (IoT).