An Integrated Optimal Method for Cloud Service Ranking

2021 Turkish Journal of Electrical Engineering and Computer Sciences  
Many cloud providers present various services with different attributes. It is a complex, lengthy process to 3 select a cloud service that meets user requirements from an assortment of services. At the same time, user requirements 4 are sometimes defined with imprecision (sets or intervals), whereas it is also important to consider the Quality of User 5 feedback (QoU) and Quality of Service (QoS) attributes for ranking. Besides, each MADM method has a different 6 procedure, which causes
more » ... al contradictions. These contradictions have led to confusion in choosing the best 7 MADM method. It is necessary to provide a method that harmonizes the results. Therefore, choosing a method for 8 ranking cloud services that addresses these issues is currently a challenge. This paper proposes an Optimal Cloud Service 9 Ranking (OCSR) method that ranks cloud services efficiently based on imprecise user requirements in both QoS and 10 QoU aspects. OCSR consists of four stages including receiving the requirements, preprocessing, ranking, and integrating 11 the ranking results. At the receiving requirements stage, the query format is created. In the preprocessing stage, 12 a requirement interval is created for considering imprecise user requirements in order to filter inappropriate services. 13 Based on QoS and QoU attributes, cloud services are then ranked through multiple Multi-Attribute Decision-Making 14 (multi-MADM) methods such as the prominent MADM techniques. Finally, the ranking outputs of various methods 15 are integrated to obtain the optimal results. The experimental results confirm that the OCSR outperforms the previous 16 methods in terms of optimality of ranking, sensitivity analyses, and scalability. 17 Key words: Cloud service selection, Cloud service ranking, Multi-attribute decision-making (MADM), Quality of service 18 (QoS), Quality of user feedback (QoU) 19 1. Introduction 20 Cloud computing plays a prominent role in developing systems and distributing applications on the internet. 21 Cloud provides easy, safe, flexible, and scalable access to information and computational resources (infrastruc-22 ture, platform, and software) in various services [1]. There are many cloud computing providers that offer 23 similar services with different qualities. At the same time, cloud users always have different requirements for 24 various applications, a fact which makes it complex and time-consuming to compare and select cloud services 25 meeting user requirements [2]. 26 With the development of cloud computing and the availability of abundant services, it is challenging to 27 devise an appropriate method for selecting and ranking services [3]. User requirements are also often met by 28 more than one service; therefore, ranking systems are employed to provide the most appropriate service (i.e. the 29 highest-quality service meeting user requirements) at short notice [4]. Basically, ranking cloud services would 30 * Correspondence: motameni@iausari.ac.ir This work is licensed under a Creative Commons Attribution 4.0 International License. necessitate considering two perspectives: quality of service (QoS) and quality of user feedback (QoU) [2]. There 1 are several contradictory attributes in each perspective [5]; such as achieving high CPU speed and availability 2 apart from minimizing latency and price in QoS aspect. Also, achieving high security and trust apart from 3 minimizing response time in QoU aspect. In addition to the contradictory quality attributes, the different 4 requirements of users, a large number of cloud services, and user feedback from cloud services are effective in 5 ranking services. Therefore, these factors have made ranking cloud services a complex issue [5]. To address 6 these problems, Multi-Attribute Decision-Making (MADM) methods can be considered as the best choice [6]. 7 The MADM methods can help users to simplify and solve the problem of ranking cloud services [1]. 8 Researchers have proposed various methods for ranking cloud services. Many studies address QoS 9 attributes, and extensive efforts have been made to identify the metrics for measuring these attributes in cloud 10 computing environment [7-9]. Quite a few researchers have focused on user requirements and proposed methods 11 for measuring services based on requirements [1, 10, 11]. However, some researchers have evaluated cloud services 12 in terms of trustworthiness and security perspective [3, 12], and many others have proposed solutions to the 13 improvement of existing methods [5, 13, 14]. For instance, Garg et al. proposed a ranking cloud services 14 framework based on user requirements [1]. Because user requirements can be imprecise, solutions to improve 15 this method have been proposed. For this purpose, Kumar et al. proposed a method of combining MADM 16 methods and fuzzy to handle imprecise in the user requirements [13]. The combination of these methods increases 17 the response time. To reduce the response time, a method was designed by applying changes in combination 18 of MADM methods [14]. As fuzzy sets have limitations for considering imprecise user requirements, some 19 parameters were added to the fuzzy set to increase the ranking trustworthiness [5]. 20 Recent cloud service ranking methods have reported similar deficiencies. Nearly 100 quality attributes 21 have been proposed so far, and each study uses only a few of them [5]. Most of the recent methods have 22 overlooked user feedback when ranking services [4], whereas many have used different ranking methods yielding 23 different results [2]. Furthermore, the complex and lengthy comparisons have prevented them from responding 24 when the number of cloud services is high [3]. However, studies have often neglected to rank cloud services within 25 a framework to (1) consider optimality (the highest level of proportion to user requirements) and execution time; 26 (2) perform ranking based on imprecise (sets or intervals) user requirements; (3) provide an approach to merge 27 QoU and QoS aspects. Fixing these deficiencies can effectively improve the ranking of cloud services. 28 This study presents an integrated method for ranking cloud services to perform ranking of high optimality 29 and reduce execution time. In the OCSR, ranking is based on user requirements to consider QoS and QoU 30 attributes. To reduce execution time, OCSR proposes an approach to eliminate services that do not meet 31 the user requirements. The OCSR method uses multi-MADM method to rank cloud services. Multi-MADM 32 method includes prominent MADM methods and uses their ranking results to achieve optimal rankings. The 33 reason for using multi-MADM methods is that each MADM method has a different procedure that leads to 34 contradictory rankings. Such contradictions in the ranking of cloud services cause confusion in selecting the best 35 service. Therefore, the OCSR has provided a method to integrate contradictory ranks of cloud services obtained 36 from each of the MADM methods. The OCSR method has been assessed in three scenarios such as optimality, 37 sensitivity analysis and scalability. The results show that the OCSR method retained an acceptable execution 38 time by using multi-MADM for ranking. At the same time, it shows a higher optimality and sensitivity rate 39 than other methods. 40 The main contributions of this research are summarized as follows. 41 • An integrated method called OCSR based on imprecise user requirements has been proposed. OCSR 42 considers user's QoS and QoU requirements to provide a broader view of service rankings. 1 • OCSR emphasizes preprocessing to improve the results in the following ways. (1) Appropriate services 2 are selected in the preprocessing stage, and a limited number of services are ranked. (2) Services are 3 not eliminated due to non-compliance with one or multiple user requirement attributes. Instead, service 4 evaluation is based on all QoS and QoU attributes. (3) Preprocessing is automated and needs no expert 5 involvement. 6 • OCSR employs the multi-MADM to rank cloud services. The ranking results obtained from each MADM 7 method are integrated using a proposed method to produce consensual ranking results. 8 • A comparison has been applied based on optimality, sensitivity, and scalability metrics between OCSR 9 and other existing methods. 10
doi:10.3906/elk-2010-54 fatcat:zhybkb35uba7hfjnzkyxhbvjie