Green cloud computing adoption challenges and practices: a client’s perspective-based empirical investigation

Over the last decade, the widespread adoption of cloud computing has spawned a new branch of the computing industry known as green cloud computing. Cloud computing is improving, and data centers are increasing at regular frequencies to meet the demands of users. On the other hand, cloud providers pose major environmental risks because massive data centers use a large amount of energy and leave a carbon footprint. One possible solution to this issue is the use of green cloud computing. However, clients face significant difficulties in adopting green cloud computing. This study aims to understand the problems faced by client organizations while considering green cloud computing. In addition, this study aims to empirically identify the solution to the challenges faced by green cloud computing practitioners. A questionnaire survey approach was used to get insight into green cloud computing practitioners concerning the challenges they faced and their solutions. Data were obtained from sixty-nine professionals in green cloud computing. The results revealed that “lack of quality of service”, “lack of dynamic response”, and “lack of services to satisfy client’s requirements” are critical for green cloud computing. In addition, sixty-three practices for addressing the challenges in green cloud computing are also identified. The identified challenges and practices of green cloud computing will benefit the client organizations to update and revise their process to consider green cloud computing. In addition, it will also assist vendor organizations in developing, planning, and managing systems concerning client satisfaction.


Introduction
Green cloud computing (GCC) has grown in popularity over the past decade. Owing to the widespread use of cloud computing, the cloud computing approach is rapidly gaining traction. Cloud computing (CC) providers regularly expand their data centers [1]. CC is a backbone infrastructure to satisfy the constantly expanding computational and storage demands [2]. CC demands increase daily due to its 24/7-hour availability and cheaper services. Owing to the increasing demand for CC services, the result of digital transformation, and the cloud's high exibility, more efforts are needed to increase the energy effectiveness of systems in cloud data centers [3]. Demand and competition cause cloud companies to be put under constant pressure, to decrease costs while improving the quality of their services [4]. Thereby, services are no longer provided from a single place but a global pool of resources spread across heterogeneous ecosystems [4][5][6].
Customers can use CC resources in virtual machines (VMs) installed and ran within data centers. The data centers are composed of several physical servers, each with its resources. As a result, each cloud has a signi cant number of services that require a large amount of electricity, increasing CO2 emissions [7]. Jones [6] predicted that by 2030, information and communication technology activities would account for 20.9% of the world's electricity demand. Based on the above observations, it is understood that cloud providers face a severe challenge in making clouds environmentally friendly. Additionally, the need for new areas is inevitable. To cope with environmental demand, a new area, known as Green Cloud Computing (GCC), was introduced in the cloud computing domain.
GCC comprises two terms: green, which signi es environmentally bene cial, and cloud, which is a typical internet symbol [8]. Therefore, it can be deduced that GCC is the eco-aware use of resources to make computers environment-friendly. Many IT businesses are now taking the lead in developing energy-e cient computing solutions that are also recyclable, which decreases the use of hazardous materials [9]. Green computing reduces CO2 emissions by allowing clients to use web services to request resources and pay for those used [10,11].
GCC can be tackled from several perspectives, including hardware, software, legislative, administrative, and executive challenges [12][13][14]. Consequently, green IT advancement in data centers has become critical [6, 10], with the green cloud below the mentioned bene ts: To make sure energy bills are the least.
To enhance the company's productivity.
To make lifelong equipment.
To enhance reliability and limit the downtime risk.
Fewer carbon emissions.
Despite the advantages of green cloud computing, GCC faces numerous hurdles. Many scholars have identi ed the general di culties and practices of GCC. The following are some of the challenges faced by GCC [15]: Resource allocation is dynamic and based on energy.
Resource allocation and selection are based on the quality of service.

Optimization of virtual network topologies
Automatic optimization of a device's heat condition and freezing operation.
Effective VM aggregation for diverse workload management.
Even though signi cant effort has been made in GCC [5] to identify di culties and associated practices, they have primarily focused on a broad perspective. Similarly, a literature review on incorporating green cloud attributes and the vendor's genuine employment has been published from the vendor's perspective [2].
However, there is no literature study from the client perspective in GCC to provide conceivable solutions or practices to clients.
Hence, this study gap motivated us to explore challenges and practices from the client's perspective in GCC. Clients are important to businesses in one manner or another. To ensure that GCC is successful in terms of client satisfaction level, we need to validate the challenges and practices (to these challenges), we identi ed in our previous study using SLR [1]. We also intended to identify new challenges other than the identi ed ones and to nd new practices from industry practitioners. To accomplish the aims presented above, we have articulated the following research questions (RQs) for this empirical study: The rest of this paper is arranged as follows: Section 2 shows the background of green cloud computing. Section 3 describes the research technique in full. Section 4 provides our survey results. The study's limitations are described in section 5 and the conclusion and future work in Section 6. Additional information is provided in the remaining sections.

Background
The computing business has progressed more rapidly than any other science discipline, from the classical to the modern age. It is also worth noting that the employment of evolving computers has enabled a variety from the classical to the modern age. Computers are used in every aspect of science, either directly or indirectly. To keep up with rising demand, the computer sector has been rapidly expanding, causing environmental damage. We cannot stop the expansion of computer systems because we require them to meet our demands. As a result, we require a solution to this problem. Green cloud computing has been the subject of substantial research in the computer industry over the last decade. The following are some examples of noteworthy work completed.
Joy et al. [16] compared service level agreements (SLAs); because it is so complex, it is much harder to identify the fundamental cause of service disruption in the cloud infrastructure. Goyal et al. [17] highlighted that current cloud computing infrastructures must provide a high quality of service (QoS) to their clients, necessitating a trade-off between power and performance. Federico et al.
[18] used a mixed-integer linear We conducted a survey of clients in order to assess the challenges associated with adopting GCC and gather information on potential solutions. Surveys are a research method used to gather tacit knowledge on a particular topic by asking questions and collecting data in order to obtain a quantitative or numerical description of a speci c population. Other researchers [26][27][28][29][30] have taken a similar strategy. We conducted a survey via an online questionnaire created using Google Forms "https://forms.gle/4bsjrvFQaZJvumb9A", a free web tool. The main objective of the questionnaire survey was to validate our SLR ndings through industry practitioners, discover new and unresolved challenges not mentioned in the previous study, and nd out about best practices for each emerging challenge. The following sections discuss the steps involved in conducting the questionnaire survey.

Empirical Study
A questionnaire survey was used to gather data. It was necessary to follow the safety guidelines for the COVID-19 pandemic. Thus, to continue our research while adhering to the SOPs, we opted to distribute the questionnaire survey via the Internet. We initially identi ed relevant groups using web forums, such as LinkedIn. A consent form was posted in each group. Table 1 contains information on the number of members in each category. A numeric description of the sample is assigned to the empirical survey strategy to get implicit knowledge of a precise or perplexing occurrence. Other researchers commonly employ the empirical survey to collect relevant data [37][38][39][40][41].

Designing Online Questionnaire Survey
Our survey involved various steps, including sampling, selection of appropriate individuals, and questionnaire design. Sampling is the process of selecting a representative subset of individuals from a larger population to participate in the survey by completing the questionnaire [25]. To gather the necessary information for our survey, we utilized a questionnaire strategy, which involved the creation of a set of questions for participants to answer. The subsequent sections will delve into further detail on both of these aspects.

Sampling
There are two methods of sampling: systematic and non-systematic approaches [25]. Since the complete population list was readily available for a small-scale survey, a non-systematic approach was utilized where samples were selected from the list based on statistical analysis [25]. We have used non-systemic methods to survey contacts of a big number of companies that are di cult to get. Similar methods were used by other researchers [37,38,40,42,43] as well. In our non-systematic sampling approach, we customized the process according to our needs. We initiated the dissemination of questionnaires by creating an invitation letter summarizing our research and sent it out to various websites.
3. We invited all the authors of our selected SLR papers because they are relevant to the speci c area. Furthermore, they can contribute more to the research due to their relevance with the area. The authors

Data Analysis
Upon receiving the questionnaires, we carefully reviewed them and excluded any low-quality responses that did not meet our pre-de ned criteria. These criteria included the following: Respondents with a level of education lower than a bachelor's degree.
Incomplete or partially lled questionnaires.

Using our secret questions
By applying these criteria, we ensured the quality and reliability of the survey results.
We received a total of 88 responses from participants around the world. However, after applying the quality criteria, 19 responses were excluded. The remaining 69 responses were carefully evaluated and used for data analysis.

Results And Discussion
In this section, we present the outcomes of our research concerning the research question mentioned in Section 1. Table 2 presents the ndings of our empirical research, which outlines the issues discovered in response to RQ1. Based on the ndings, six of the nine challenges have > 50% frequency, while the remaining three have a 48% frequency of occurrence. Our research revealed that the most frequent issue faced by users is the "lack of service quality," which occurred 75% of the time. This highlights the importance of ensuring that services provided by green cloud computing vendors meet users' expectations in terms of quality.

Challenges identi ed through empirical study
Additionally, our ndings indicated that the "lack of dynamic response" was the second most common issue, accounting for 72% of the occurrences. Users seek assurance that the services offered dynamically, during both high and low-demand periods, remain unaffected. Hence, green cloud computing vendors need to ensure that their services are not impacted by load uctuations.
Our research also revealed that the third most common issue faced by users is the "lack of services to ful ll client's speci c requirements," which occurred 64% of the time. This underscores the importance of vendors meeting their users' speci c needs and expectations as per the contract, which is crucial in building a trustworthy relationship between the vendor and the client. Other frequently mentioned positive challenges include: "lack of green and sustainable infrastructures for the provision of reliable and cheaper services" (55%), "High cost" (52%), and "lack of e cient provision of resources and services" (52%). The results also indicate that out of nine challenges, the "lack of e cient provision of resources and services' challenge has a 35% occurrence in the unfavorable list of Table 2.The second highest unfavorable responses for the challenge "Lack of Green and sustainable infrastructures for the provision of reliable and cheaper services" had 30 responses. The third highest unfavorable responses received for the challenge "High cost," has 28 responses.
In the neutral list in Table 2, all the challenges have been cited. This indicates that these practitioners have little knowledge of these challenges. The rst most cited challenge on the neutral list is "lack of cybersecurity," which received 35% of the total. The second most cited challenge is "SLA violation," which received 29% of the total, and the third most cited challenge is "complexity in understanding cloud systems," which received 27% of the total. According to the ndings, six of the nine challenges had a favorable list occurrence rate of greater than 50%. Table 2 shows that the remaining three had a 48% favorable list occurrence. We did not nd any challenge with an occurrence greater than 50% in the unfavorable list.
The criteria are F = Favorable, UF = Unfavorable, and N = Neutral ( Table 2). The F includes results of strongly agree (SA) and agree (A), and UF includes strongly disagree (SD) and disagree (D). The responses that neither agree nor disagree are considered in the N = Neutral category. We conducted an online survey in two steps. First, we disseminated our survey using different online groups: Linkedln and WhatsApp. We received a total of 38 responses; applying our quality criteria, the responses were reduced to 27. We used the verr platform to disseminate our survey in the second step. We received a total of 50 responses, and based on our quality criteria presented in section 3.3, it was reduced to 42. We performed two different analyses on these different platforms and analyzed their differences.

Analysis based on a platform used other than Fiverr
We consider the challenges critical if their occurrence is greater than 50%. The participants considered seven critical challenges, as shown in Table 3. The rst most critical challenge considered is "Complexity in understanding cloud systems/ unawareness of people to use it" having a favorable percentage of 74. The second most favorable critical challenge is "Lack of quality of service" and "Lack of cybersecurity" with 67%.
The third most favorable critical challenge is "Lack of services to address clients' speci c requirements". The most unfavorable challenge is the lack of e cient provision of resources and services, with 30%.

Analysis based on the Fiverr platform
The participants considered six critical challenges, as shown in Table 4. The most favorable critical challenge considered by the participant is "Lack Of quality of service" and "Lack of dynamic response," having a favorable percentage of 81. The second most critical challenge is "Lack of services to address client's speci c requirements," with 67%. The third most critical challenge is the lack of e cient provision of resources and services, with 55%. The participant also considered the most unfavorable challenge is "Lack of e cient provision of resources and services," having a percentage of 38. The second most unfavorable challenge considered is the "Lack of Green and sustainable infrastructures for the provision of reliable and cheaper services" and "Complexity in understanding cloud systems/ Unawareness of people to use it. Further analysis in Table 5 shows that "lack of cybersecurity" and "complexity in comprehending cloud system" are the high risks for large-sized organizations with 75% and should be treated on a priority basis.
The second major obstacle that large enterprises are considering is "a lack of service quality" and "a lack of green and sustainable infrastructure for the provision of reliable and cheaper services", with 67%, as shown in Table 5. Other four challenges with percentage > 50 are: "a lack of the dynamic response", "lack of services to suit client's special requirements", "high cost", "SLA violation," and "strategies should be de ned to control them". Our ndings indicate that medium-sized companies consider "lack of quality of service" a signi cant challenge client's face in adopting GCC 86%. The second most critical challenge is the lack of a dynamic response (83%). The third most critical challenge is "high cost," with 72%. However, results indicate that for small companies, "lack of dynamic response" is the critical challenge faced by the client (71%). The second most critical challenges are "lack of quality of service" and "lack of services to address client's speci c requirements," having a percentage of 68%, and the third most critical challenge is "lack of Green and sustainable infrastructures for the provision of reliable and cheaper services" having a percentage of 64%. Figure 1 shows the graphical representation of the positive responses received from different organization sizes.

Challenges responses based on the continent
We have received responses from all over the world. The most responses we received were from China 11, and the second most responses that we received were from Pakistan 10. Figure 2, the responses received are divided into four major groupings: Asia, Europe, North America, and other regions. Other refers to countries not located in Asia, Europe, or North America. Other countries include South Africa, New Zealand, Australia, Brazil, and Peru. The continent of Asia had the most responses, accounting for 39% of all responses. The other category, which accounts for 25% of the total, is the second. With 22%, North America is the continent with the third most responses, and from Europe, we received 14% of the total responses. From Table 6,"Lack of quality of service" has been considered the most critical challenge: having a frequency of 52% in the favorable occurrences. "Lack of dynamic response" has been considered the second most-critical challenge: having a frequency of 50% in the favorable occurrences. "Lack of services to address client's speci c requirements" has been considered the third most critical challenge: having a frequency of 44% in the favorable occurrences. "Lack of e cient provision of resources and services" has been considered the most unfavorable challenge, with a frequency of 24%. Likewise, "Lack of Green and sustainable infrastructures for the provision of reliable and cheaper services" has been considered the second most unfavorable challenge with a frequency of 21.
We also used a linear by linear Chi square test to see whether there was any signi cant difference between challenges across the different continents. According to our ndings, the "complexity in understanding cloud systems" challenge signi cantly differs from a linear by linear Chi square value of 0.007, which is lower than the prede ned value, as shown in Table 6. Other than this challenge, there is no discernible change. We may say that different continents have similar perspectives on the problems identi ed. We can generalize it because users worldwide face the same problems.

Empirical analysis based on company care about green computing
Our survey included a question regarding the companies' concern about green computing. The participants were given the options of Yes, No, and Maybe to respond to this question, and their answers were analyzed and presented in Table 7. Based on our ndings, the majority of the participant companies expressed concern about green computing, with only 47.1% of organizations indicating that they do not care about green computing.
On the other hand, 8 participants are unsure if their company is concerned about green computing. "Lack of quality of service" has been considered the highest challenge by the participants, whose company cares about green computing, having an occurrence of 44% in the favorable list. "Lack of dynamic response" has been considered the second-highest challenge in the favorable list, with an occurrence of 42%. "SLA violation" and "Complexity in understanding cloud systems" have been considered the highest challenges in the unfavorable list, with a frequency of 23%. We performed an LBL-CST to analyze differences in the company's approach to green computing, and found that "Lack of quality of service" had a signi cant difference with a Chi-square value of 0.001, below the prede ned threshold of 0.05. Similarly, "Lack of dynamic response" also shows a signi cant difference having an LBL-CST value of .009, which is also less than the prede ned value. We highlighted it in Table 7 for easy understanding. The other seven challenges show no signi cant difference Fig. 3 presents a graphical representation of responses received from participants.

Practices identify through empirical study
We identi ed 63practices for the nine challenges. The responses received for each practice are shown in Table 8 to Table 16.        to provide an application's computational resources on demand. Seeking the signi cant returns of GCC, this study empirically explores the challenges practitioners face while considering GCC in practice. The results render that "lack of quality of service", "lack of dynamic response", and "lack of services to satisfy client's requirements" are critical for GCC. In response, one of the practitioners also reported an additional challenge of GCC, i.e., "lack of trust".
We further analyzed the identi ed GCC challenges in context to client organizations based on the platform used by Fiverr, the platform used other than Fiverr, the company's size, continents, and the importance the company gives to GCC.
The analysis based on the platform used other than Fiverr represents that the top-most critical favorable challenges facing GCC client organizations are "complexity in understanding cloud systems/unawareness of people to use it", "lack of quality of service", "lack of cybersecurity" and "lack of services to address clients' speci c requirements".
The analysis based on the platform used by Fiverr represents that the top-most critical favorable challenges facing GCC client organizations are "lack of quality of service" and "lack of dynamic response", "lack of services to address client's speci c requirements," and "the lack of e cient provision of resources and services". The participant also considered the most unfavorable challenges are "Lack of e cient provision of resources and services", "lack of green and sustainable infrastructures for the provision of reliable and cheaper services," and "complexity in understanding cloud systems/ unawareness of people to use it".
Further analysis in Table 5 shows that "lack of cybersecurity" and "complexity in comprehending cloud systems" are high risks for large organizations. Our ndings indicate that medium-sized companies consider "lack of quality of service", "lack of a dynamic response," and "high cost,". However, results indicate that for small companies, "lack of dynamic response", "lack of quality of service", "lack of services to address client's speci c requirements," "lack of Green and sustainable infrastructures for the provision of reliable and cheaper services" are the critical challenge faced by the GCC client organizations.
We utilized a linear by linear Chi square test to determine if there were any notable differences between the challenges occurred on different continents. According to our ndings, the "complexity in understanding cloud systems" challenge signi cantly differs from a linear by linear Chi square value of .007, which is lower than the prede ned value, as shown in Table 6. Other than this challenge, there is no discernible change. We may say that different continents have similar perspectives on the problems identi ed. We can generalize it because users worldwide face the same problems.
According to our ndings, most participant companies care about green computing, with 52.9% of organizations not caring about green computing. On the other hand, 8 participants are unsure if their company is concerned about green computing or not. "Lack of quality of service" has been considered the highest challenge by the participants, whose company cares about green computing, having an occurrence of 44% in the favorable list. "Lack of dynamic response" has been identi ed as the second most signi cant challenge on the list of favorable, with a frequency of occurrence of 42%. "SLA violation" and "complexity in understanding cloud systems" have been considered the highest challenges in the unfavorable list, with frequency of 23%.
We also used an LBL-CST to see if there was a substantial difference between the company's approaches to green computing. We found that "lack of quality of service" shows a signi cant difference having a Chisquare value of 0.001, which is less than the prede ned value of .05. Similarly, "lack of dynamic response" also shows a signi cant difference having an LBL-CST value of .009, which is also less than the prede ned value.
In future work, we also plan to identify success factors related to GCC from the client perspective and validate them from the industry. We will also develop a roadmap based on challenges, success factors, and their associated best practices. This roadmap will assist vendors in focusing on areas that need further development and planning to improve client satisfaction levels while adopting GCC.

For Researchers
This research study has important implications for academic research in several ways. First, in terms of problem de nition, we extensively examined the obstacles that clients face when adopting GCC. Second, regarding the solution and presentation, we systematically studied and analyzed the current literature on GCC, and validated it with real-world industry data. Third, our research has con rmed that the successful implementation of GCC activities can result in organizational success. Finally, this study provides a comprehensive knowledge base of green cloud computing from the client's perspective, which has not been previously achieved.

Implications for industry
Our study's results have practical implications for practitioners in several ways. Firstly, practitioners can use the list of identi ed challenges and practices as a knowledge base to implement GCC in an organization. Secondly, organizations can utilize the identi ed challenges to improve their systems by developing, planning, and managing the requirements that meet clients' needs. Thirdly, practitioners can bene t from focusing on the list of critical challenges and 63 practices identi ed in the study (Table 8-16) to more effectively implement GCC in an organization. Additionally, organizations can use the list of critical challenges and their mapped practices as an indicator to hire cloud computing experts with speci c skills/practices as a risk mitigation strategy for GCC platforms. The study ndings can also assist vendor organizations in GCC in setting better plans and strategies that meet clients' demands. Ultimately, from a client perspective, this study will help organizations nd solutions to the risks encountered while adopting GCC.

Study Limitations
In this section, we elaborate on the potential threats and how they were mitigated following the criteria suggested by [44][45][46].

External validity -Generalizability
This study is based on empirical data collected from a speci c context, namely GCC clients. Therefore, our ndings do not purport to be generalizable to all other GCC contexts. Instead, this study aimed to conduct a comprehensive investigation of the challenges faced by GCC clients to provide detailed explanations through rigorous data analysis. However, we do not claim that the results are absolute or nal. Instead, they can be replicated and adapted to other contexts [47,48].  Challenges based on company approach towards GCC adoption