Cloud computing (CC) has recently been receiving tremendous attention from the IT industry and academic researchers. CC leverages its unique services to cloud customers in a pay-as-you-go, anytime, anywhere manner. Cloud services provide dynamically scalable services through the Internet on demand. Therefore, service provisioning plays a key role in CC. The cloud customer must be able to select appropriate services according to his or her needs. Several approaches have been proposed to solve the service selection problem, including multicriteria decision analysis (MCDA). MCDA enables the user to choose from among a number of available choices. In this paper, we analyze the application of MCDA to service selection in CC. We identify and synthesize several MCDA techniques and provide a comprehensive analysis of this technology for general readers. In addition, we present a taxonomy derived from a survey of the current literature. Finally, we highlight several state-of-the-art practical aspects of MCDA implementation in cloud computing service selection. The contributions of this study are four-fold: (a) focusing on the state-of-the-art MCDA techniques, (b) highlighting the comparative analysis and suitability of several MCDA methods, (c) presenting a taxonomy through extensive literature review, and (d) analyzing and summarizing the cloud computing service selections in different scenarios.
Cloud computing (CC) is the distributed computing model which provides computing facilities and resources to the users in an on-demand pay-as-you-go model [
Service provisioning is an important aspect of CC because it directly impacts the user experience of the service. New users and companies seeking cloud services are constantly emerging. The vast diversity among the available cloud services makes it difficult for the customer to decide whose services to use or even to determine a valid basis for their selection [
Over the past few decades, decision-making theory has been successfully applied in a growing number of diverse domains and has assisted in decision making, including several well-known examples. The multicriteria decision-making (MCDM) approach is capable of handling multiple conflicting criteria [
One objective of this study is to survey the literature and to provide a critical assessment of the available MCDA techniques and their usage in service selection for cloud computing. A literature review is used to demonstrate the integration of MCDA techniques and cloud computing based on their usage and popularity. Hence, in this paper, we reviewed the current literature and identified the different types of problems. The limitations of the various methods and techniques and their pros and cons are also discussed. We highlight the fact that most MCDA techniques have often been used individually in previous studies. Finally, we provide crucial information through reviewing the available literature on MCDA techniques and its usage in service selections in cloud computing [
Hence, we identify our contribution of this review are four-fold: (a) focusing on the state-of-the-art MCDA techniques, (b) highlighting the comparative analysis and suitability study of several MCDA methods, (c) presenting a taxonomy through extensive literature review, and (d) analyzing and summarizing the cloud computing service selections in different practical implementation aspects.
The rest of this paper is organized as follows. Section
In 1951, Harold William Kuhn and Albert William Tucker introduced the vector maximum problem, the first explicit consideration of the basic concepts of MCDA. In 1972, a conference on “Multiple Criteria Decision Making (MCDM)” was held at Columbia University in South Carolina. Since then MCDA/MCDM has experienced rapid growth and continues to grow today [
Taxonomy of MCDA.
MCDA, one of the most important branches of operations research, aims to design mathematical and computational tools for selecting the best alternative among several choices, with respect to specific criteria, either by a single decision maker or by a group. The field consists of two main categories: multiattribute decision making (MADM) and multiobjective decision making (MODM). According to a survey conducted to evaluate the use of different methods, especially for improving quality, fuzzy was the most frequently used, representing 40% of the total, followed by AHP and ANP together at approximately 20%. TOPSIS was also widely used. In contrast, DEA, Goal Programming, and ELECTRE were employed only rarely [ MCDM is a collection of methodologies for comparing, ranking, and selecting multiple alternatives, each having multiple attributes. It depends on a matrix called the evaluation matrix, decision matrix, payoff matrix, or evaluation table. MCSP selects the best alternative from a finite set of alternatives, all of which are known a priori. MCMP selects the best alternative from a very large or infinite set of alternatives, not all of which are known a priori. MAUT finds a utility function reflecting the usefulness of a particular alternative.
MCDA methods can be categorized into two types: (1) multiattribute utility theory (MAUT) and (2) outranking methods. MAUT attempts to find a function reflecting the utility or usefulness of a particular alternative. Each action is assigned a marginal utility, with a real number representing the preferability of the considered action. The returned utility is the sum of these marginal utilities. Outranking methods decide whether one alternative is ranked higher than another by employing a pairwise comparison.
MCDA methods are divided into multiobjective decision making (MODM) and multiattribute decision making (MADM). The two methods differ mainly by how the alternatives are enumerated. In MODM, they are not predetermined but arise from the optimization of a set of objective functions. In MADM, they are predetermined, and a small subset is evaluated against a set of attributes. In both methods, the best alternative is chosen by comparing the rankings of each alternative/attribute combination [
In 1980, Thomas L. Saaty discovered AHP, a popular and widely used method for MCDA. AHP allows the use of qualitative as well as quantitative criteria when evaluating alternatives and the attributes are not entirely independent of each other. The AHP is based on a pairwise comparison, with the attributes structured into a hierarchal relationship, which is very useful. The hierarchy starts from the top level towards the goal; the lower levels correspond to criteria, subcriteria, and so on. In this hierarchy tree, the process starts from leaf nodes and progresses up to the top level. Each output level represents the hierarchy corresponding to the weight or influence of different branches originating for that level. Finally, after making the comparisons, the best alternative with respect to each attribute is usually selected [
ANP is an extension of AHP proposed by Thomas L. Saaty in 1996. It is a comprehensive decision-making technique designed to overcome the problem of dependence and feedback among the criteria, using unidirectional hierarchical relationships between decision levels. ANP describes interrelationships among the decision levels and attributes using unidirectional hierarchical relationships with dependence and feedback, employing ratio scale measurements based on pairwise comparisons to model the decision problem. To handle interdependence among elements, ANP derives a “supermatrix” containing composite weights [
This technique shows preference for the similarity to an ideal solution, which tries to select an alternative that is closest to the ideal solution and simultaneously farthest from the anti-ideal solution. In this technique, the decision matrix is first normalized using vector normalization, and the ideal and anti-ideal solutions are identified within the normalized decision matrix.
TOPSIS was developed in 1981 by Hwang and Yoon, and it selects alternatives having the shortest distance from the positive ideal solution and the farthest distance from the negative ideal solution [
ELECTRE was developed in 1991 by Roy and colleagues at the SEMA Consultancy Company. Several versions of this method have been developed since (ELECTRE I, ELECTRE II, ELECTRE III, ELECTRE IV, ELECTRE IS, and ELECTRE TRI (ELECTRE Tree)). ELECTRE consists of two sets of parameters: the importance coefficient and the veto thresholds [
This method falls in the class of outranking MCDM methods. In comparison with the previously discussed methods, this method is computationally intricate: the simplest variant of ELECTRE involves up to 10 steps. It performs a pairwise comparison between the alternatives in order to determine their outranking relationships. These relationships are then used to identify and eliminate alternatives that are dominated by others, yielding a smaller set of alternatives.
The ELECTRE method handles discrete criteria that are both quantitative and qualitative in nature, providing complete ordering of the alternatives. Alternatives are preferred over most of the criteria and depend on concordance, discordance indices, and threshold values and graphs of relationships. These graphs are used in an iterative procedure to obtain the ranking of alternatives [
PROMETHEE was developed in the mid-1980s by Brans and Vincke. It is an improved form of the outranking method ELECTRE, but it differs from ELECTRE in the pairwise comparison stage. Whereas both PROMETHEE and ELECTRE determine whether one particular alternative is better than another, PROMETHEE additionally considers the degree to which it is better, using this piece of information to eliminate dominated alternatives and to identify nondominated or least dominated alternatives. PROMETHEE ranks the alternatives and is easier to use and less complex than ELECTRE [
Gabus & Fontela developed DEMATEL in 1973 at the Geneva Research Centre of the Battelle Memorial Institute. This method represents factors as interrelationships among criteria. Hence, DEMATEL is a complete method for building a structural model involving associations of complex factors. It organizes relationships between the elements within a system using numerical representations of the power of influence [
The grey system theory proposed by Deng (1982) has been widely applied in many fields. The term “grey” interpreted as a color is intended to suggest the amount of known information in control theory. GRA, derived from grey system theory, is particularly useful when dealing with poor, incomplete, and uncertain information. It is suitable for solving problems with complex interrelationships between factors and variables, and it has been successfully applied to solve a variety of MADM problems. The main advantages of GRA are the fact that the results are based on original data and that calculation is straightforward and simple [
VIKOR, also known as the compromise ranking method, is an effective tool in multicriteria decision making. The acronym is derived from the word ViseKriterijumska Optimizacija I Kompromisno Resenje. Its multicriteria ranking index is based on a measure of “closeness” to the “ideal” solution. This method was introduced by Opricovic in 2004 for optimization and compromise evaluation in dynamic and complex processes.
VIKOR employs linear normalization, but the normalized values do not depend upon the evaluation unit of a criterion. An aggregating function balances the distance from the ideal solution between individual and ideal satisfaction [
In 1965, Zadeh proposed fuzzy set theory, which has been extensively applied to model the ambiguities of human judgment. It also effectively resolves uncertainties in available information for multiple criteria decision making. A fuzzy MCDA model is used to evaluate selected alternative criteria by using decision pools. The suitability of replacements versus criteria and the significance weights of criteria are evaluated in terms of linguistic values represented by fuzzy numbers [
In 1955, Charnes first employed goal programming as an MODM tool. Goal programming is an extension of linear programming used to solve problems containing multiple, usually conflicting, objects. It is an optimization procedure for handling multiple conflicting objective measures. It is widely used in multicriteria decision making to combine the logic of optimization with mathematical programming in order to make decisions fulfilling several objectives [
Data envelopment analysis (DEA) is a mathematical programming technique used for evaluating the competence of an observation relative to a set of similar observations [
Summary of MCDA techniques and capabilities.
Name | Objective | Criteria/approach | Author and year |
---|---|---|---|
Goal programming | Application of linear programming to solve problems relating to multiple and conflicting objects | Combination of the logic of optimization with mathematical programming | Charnes et al. (1955) |
|
|||
Fuzzy | Evaluation of significance weights in terms of linguistic values represented by fuzzy numbers | Linguistic variables used to describe fuzzy terms that are then mapped to numerical variables | Zadeh (1965) |
|
|||
DEMATEL | Construction of a structural model involving associations of complex factors | Numerical contextual relations among the elements representing the power of influence | Gabus and Fontela (1973) |
|
|||
DEA | Evaluation of the competence of an observation relative to a set of similar observations | Mathematical programming | Charnes (1978) |
|
|||
AHP | Pairwise comparison of attributes structured in a hierarchal relationship | Useful technique for hierarchical relationship criteria | Thomas L. Saaty (1980) |
|
|||
PROMETHEE | Similar to ELECTRE but differing in the pairwise comparison stage | Considers the degree to which one alternative differs from another | Brans and Vincke (1980) |
|
|||
TOPSIS | Selection of an alternative simultaneously the closest to the ideal solution and the farthest from the anti-ideal solution | Close to ideal but the farthest from anti-ideal | Hwang and Yoon (1981) |
|
|||
GRA | Solution of problems with complex interrelationships between factors and variables | Based on grey system theory | Deng (1982) |
|
|||
ELECTRE | Pairwise comparison among alternatives used to identify and eliminate alternatives dominated by other alternatives | Checks only whether one alternative is better or worse than the other | Roy (1991) |
|
|||
ANP | More general representation of interrelationships among decision levels and attributes | Unidirectional relationships with dependence and feedback instead of hierarchy | Thomas L. Saaty (1996) |
|
|||
VIKOR | Ranking of compromises representing indices derived from a measure of “closeness” to the “ideal” solution | Employs linear normalization | Opricovic (2004) |
MCDA is a well-established area within the field of operations research, and it has proven its effectiveness in addressing different complex real-world decision-making problems. In [
AHP was proposed in order to solve this problem by using a hierarchical structure and systematization [
AHP is used by decision makers in order to make more informed decisions regarding their investment in various technologies. AHP is a multiobjective, multicriteria decision-making approach, employing pairwise comparison to derive a range of preferences from a set of alternatives. This is achieved by determining the responsiveness of the selection process to rapidly changing business rules and criteria. AHP transforms the decision-making process from a subjective judgment into an objective determination. A sophisticated formal mathematical decision model supporting the selection of cloud computing services in a multisourcing scenario is presented in [
A model for applying AHP to task-oriented resource allocation in a cloud computing environment is proposed in [
In [
Although AHP is an effective decision-making tool, disadvantages such as decision maker subjectivity can yield uncertainties when determining pairwise comparisons. In [
Although fuzzy set theory can be applied to decision-making problems possessing a degree of uncertainty, the resulting subjective judgment is always somewhat vague. Fuzzy AHP is applied to compute the fuzzy weights of each criterion based on intervalued fuzzy sets (IVFs). In [
In MCC, one of the aims is to select the optimal cloud path between certain classes of clouds that provide the same service in order to offload particular computation tasks [
Cloud service selection is a multiple criteria group decision-making (MCDM) problem. The technique for order preference by similarity (OPS) to an ideal solution (TOPSIS) can assist service consumers and providers by analyzing available services using fuzzy opinions. In [
Web services are tremendously interactive software components that can be published, located, and invoked practically anywhere on the Web. The increasing number of Web services available raises new challenges related to service discovery, selection, and composition. Machine-readable rich representations of service properties, capabilities, and characteristics can be exploited by reasoning mechanisms to support automated discovery [
TOPSIS, described in [
In [
A response time-based fuzzy control approach has been proposed for allocation of virtualized resources using a self-tuning fuzzy controller with adaptive output amplification and flexible rule selection. Based on the fuzzy controller, a two-layer QoS provisioning framework, DynaQoS was designed to support adaptive multiobjective resource allocation and service differentiation [
Summary of different applied multicriteria methods for cloud service selection.
MCDA technique | Aspects | Attributes | Reference |
---|---|---|---|
AHP | Consumer-centered service selection, especially for medical services | User preference | [ |
|
|||
TOPSIS | QoS-based multiple service selection with fuzzy options | Linguistic variable triangular fuzzy numbers | [ |
|
|||
PROMETHEE | Dynamic autonomous resource, management, and scalability | Suitable for large data centers | [ |
|
|||
AHP | Fuzzy AHP with IVFs | 2-tuple linguistic variables | [ |
|
|||
Fuzzy | Fuzzy logic-based resource evaluation technique for the DSPR framework | Fuzzy inference engine for resource evaluation. | [ |
|
|||
AHP | Identifying the scalability gain of enhanced agility in the selection process | Pairwise comparison | [ |
|
|||
Fuzzy | Response time-based fuzzy control for the allocation of virtualized cloud resources | Adaptive output amplification and flexible rule selection | [ |
|
|||
Fuzzy TOPSIS | New user centric service-oriented modeling approach in SCA. | Computational efficiency | [ |
|
|||
AHP | Decision model to support cloud computing services | Costs and risk factors | [ |
|
|||
Fuzzy DNAP and fuzzy VIKOR | Exploring interrelationships among criteria related to operations | Solves interdependence and feedback problems. | [ |
|
|||
AHP and fuzzy TOPSIS | Optimal cloud path among class of clouds to perform offloaded computation tasks | Speed, bandwidth, price, security, and availability | [ |
|
|||
AHP | Distributed resource management | Considers SLA and QoS | [ |
|
|||
ANP | QoS measuring method for cloud service architecture | A supermatrix is employed for calculation | [ |
|
|||
Fuzzy VIKOR | Assesses cloud service trustworthiness using a hybrid model | Weight-based preferences | [ |
|
|||
IVF and VIKOR | Decision analysis model for service selection | Linguistic variables | [ |
|
|||
AHP | Task-oriented resource allocation | Bandwidth, task costs, and time | [ |
In [
Evaluating the trustworthiness of a Cloud Service suffers from great uncertainty and complexity. To address this issue, a novel hybrid fuzzy multicriteria group decision-making method, based on a combination of fuzzy set and modified VIKOR methods, was proposed. It addresses various types of conflicting and incommensurable trust criteria and selects appropriate weight-based preferences in order to make a suitable decision [
A wide range of criteria is used to assess the quality of cloud computing services. A hybrid fuzzy MCDM approach combining fuzzy DEMATEL, fuzzy DANP, and fuzzy VIKOR has been used to improve service levels and meet user needs in fuzzy environments. This approach solves interdependence and feedback problems in the mobile communications industry and for related value-added service content providers by exploring interrelationships among criteria related to operations [
The selection of cloud service providers is a multicriteria decision-making (MCDM) problem. The cloud service providers with the best technology are not always suitable for a given enterprise. VIKOR is often used to solve these dilemmas. Further complicating matters, incorporating expert opinion always introduces subjectivity and vagueness to the decision-making process. Experts can use linguistic variables to express their opinions, and these variables can be used to define interval-valued fuzzy sets. A decision analysis model combining interval-valued fuzzy sets and VIKOR (IVFVIKOR) is described in [
In this study, we focus on the service selection for cloud computing in multicriteria decision-making situations. We describe the MCDA types and characteristics and present a taxonomic categorization. We compare several methods by synthesizing and reviewing the present literature. Several real-world examples with current applications of different methods are provided. Hence, MCDA has a great effect on and importance in multicriteria decision-making scenarios. We thus summarize several of the advantages and disadvantages, and we present several applications of these MCDA methods in the selection of cloud services. In addition, multicriteria applied methods are summarized and compiled in a comprehensive manner that can be applicable in other research fields. Moreover, different MCDA methods and their unique features are presented and compared, which will aid new researchers in selecting research directions. We envision that this study could be extended for intercloud service selections and for mobile cloud computing.
Analytic hierarchy process
Analytic network process
Cloud computing
Component web services
Data envelopment analysis
Decision-making trial and evaluation laboratory
Dynamic service placement and replication
Elimination and choice expressing reality
Genetic algorithm
Grey relational analysis
Intervalued fuzzy
Item-based vector similarity
Multiple attribute utility theory
Multiattribute decision making
Mobile cloud computing
Multicriteria decision analysis
Multicriteria decision making
Multicriteria mathematical problem
Multicriteria selection problem
MultiObjective decision making
Preference ranking organization method of enrichment evaluations
Quality of service
Service level agreement
Simple additive weighting
Service oriented architecture
Singular value decomposition
Technique for order preference by similarity (OPS) to an ideal solution (TOPSIS)
Universal discovery description and integration
ViseKriterijumska Optimizacija I Kompromisno Resenje
Web services
Web service composition tree.
The authors declare that there is no conflict of interests regarding the publication of this paper.
This work is fully funded by the Malaysian Ministry of Higher Education under the University of Malaya High Impact Research Grant UM.C/HIR/MOHE/FCSIT/03.