國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
The results in Table 5-3 indicate that there are three path coefficients ─ IT infrastructure, virtualization, and the relation between public and private clouds ─ with significance and having a positive relation to private cloud adoption, which supports our three hypotheses (H1-b, H1-c, H3). As for the path coefficient between service strategy and private clouds, the number 0.014 indicates significance, but it has a negative relation to private clouds. The reason behind the reverse relation may be the recent emergence of cloud computing: it is still a new concept, and most existing cloud-computing-adopted companies still adopt public and private clouds for internal use only, which causes service strategy, the concept of providing cloud services as products, to be considered less critical to readiness issues.
The results calculated using PLS support three hypotheses and reject the remainder. These results will be discussed in the follow-up section.
5.3 Findings and Discussion
We propose cloud readiness for both public and private clouds and empirically test our research model. Our analysis results assure us of some of our hypotheses but reject three of the readiness factors that we put forward. Accordingly, we summarize our findings and discussion as follows.
Finding 1: virtualization and IT infrastructure are two factors that produce positive readiness for the adoption of private clouds.
In table 5-3, the path coefficients of virtualization and IT infrastructure are significant and positive, which means that the two factors contribute to private cloud readiness when a company shifts from the absence of private cloud adoption to private cloud adoption. Regarding virtualization readiness, cloud computing involves a multitenant environment for meeting various demands from various users, which raises the issue of changing current company equipment status. Therefore, companies that are about to introduce private clouds must consider virtualization. Regarding the IT infrastructure factor, because private clouds are often defined as operating within a single organization or range, IT infrastructure built on-premise is necessary for those contemplating a move to private clouds. The results of our analysis are consistent with our hypotheses.
Finding 2: A relationship exists between public clouds and private clouds, and
companies that have adopted public clouds are more likely to succeed in the adoption
of private clouds.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
In table 5-3, the coefficient of the path between public and private clouds is highly significant, and its path coefficient is positive, which indicates positive relations in the shift from public clouds to private clouds. Private and public clouds vary according to the ownership and access rights they support, and the operation of the former is often off-limits to outsiders (Buyya, Ranjan, and Calheiros, 2010). When shifting from public clouds to private clouds, companies must begin to consider how to manage the migration and must introduce several technological changes. However, because of public cloud users’ knowledge, companies will have a better chance of adopting private clouds. The results of our analysis are in favor of our hypothesis.
Finding 3: None of the factors we proposed for public cloud readiness is significant, thus providing flimsy evidence in support of our hypotheses.
In both table 5-1 and table 5-3, the results show that all of the factors we proposed are of no significance, which means that the factors contain a high level of error and are not able to explain our research hypotheses. We summarize the possible reasons for this result below.
(1) Questionnaire-related Factors i. Small Sample Size
We reflect on the number of returned questionnaires and the increase in this number. The questionnaire was made available online for a month initially, but the number of returned questionnaires stopped increasing two and a half weeks into the questionnaire period, and we proceeded with our research. Usually, a research sample size is chosen to maximize the chance of uncovering a specific mean difference and significance among data. A larger sample is more likely to increase the precision of tests of significance between two or more reliability coefficients (Charter, 1999).
Statistical significance is a probability statement that tells us how likely it is that the observed difference is the result of chance only. The reason larger samples increase the probability of significance is that they are more likely to reach the prerequisite of significance and reliably reflect the population mean. Accordingly, our small sample size may be a partial cause of the insignificant results regarding our public cloud factors.
ii. Questionnaire Design
In addition to the small sample size used in our study, we examine the questionnaire items and attempt to scrutinize all of the item sets. To explain the
‧
phenomenon in which the mean value of PCAs (public cloud companies) are smaller than those for Non-PCAs (non-cloud-computing companies), we examine the public cloud items in particular and review them from a questionnaire recipient’s perspective.
We find that items filed under “private clouds” are more generic than those filed under “public clouds,” which indicates that private cloud item sets have greater questionnaire filler friendliness. For example, one of the items on IT infrastructure, filed under “private clouds,” is “The company is equipped with sufficient computer resources.” As for items on contract management filed under “public clouds,” one is worded as follows: “The data security of the company is not threatened by cooperation with a third party.” Because of the possible confusion caused by the item design, questionnaire recipients can be biased, thus generating the unpreferable research results.
(2) Participant-related Factors
i. Questionnaire Recipients’ Familiarity with Cloud Computing
On our questionnaire, we include several filtering items before our research question to reach our real research targets. One of the filtering questions asks recipients whether they are familiar with cloud computing and about the degree to which they are exposed to cloud computing. The question is multiple choice, and the answers vary from “I’ve hardly heard of cloud computing” to “I have years of experience with cloud computing and have instructed others in cloud computing.” In addition, we use numbers from one to five to indicate the familiarity of each questionnaire participant. The following table shows the mean value of each group.
Table 5-4. Mean Values of Cloud Familiarity of Different Groups
Non-cloud cloud companies, which is unreasonable and unsuspected. Therefore, we suggest that the mean values of group PCA in Table 5-1 are a result of the relative lack of familiarity of the private cloud users among our participants.ii. Wide Variation In Our Returned Data
‧
We inspect the returned questionnaires and search for anything that might have led to the insignificant results. First, we conduct a test on the normality of the data we retrieved, which is the focus of Table 5-5.
Table 5-5. Test of Normality
Shapiro-WilkIn Table 5-5, we use Shapiro and Wilk's W test, which is a powerful procedure for detecting normality (Royston, 1983). As a test of the normality of samples, the W statistic may be used to test normality among small samples, and it may be difficult to determine the percentage points of the distribution of a larger sample size (Shapiro and Wilk, 1965). In addition, Shapiro and Wilk’s W test rejects the null hypothesis of normality at the 0.05 level, which means that normal distribution exists when the significance number is larger than 0.05. In table 5-5, we refer to companies not using public clouds as Non-PCAs and using them as PCAs. All of the numbers in the sig.
column in Table 5-5 are greater than 0.05, which assures us of the normality of our data.
Second, we use the descriptive statistics function in the SPSS software to attempt to discover any abnormalities in our data. The results are shown in Table 5-6, which includes N as the category number, mean value, standard deviation, standard error, and minimum and maximum numbers.
‧
Table 5-6. Results of Descriptive Statistics
By scrutinizing the numbers in Table 5-6, we can observe that all of the mean values are approximately 4 and that the decimal portion of the number fluctuates.
However, the standard deviation numbers for all factors are higher than 1. In statistics, we usually use the 68-95-99.7 rule to estimate the range of a normally distributed sample. This rule is also known as the three-sigma rule, which suggests that all values of a sample lie within the range of three times the number of its standard deviation from the mean value in a normal distribution. According to this rule, if we multiply a random standard deviation number in Table 5-6 by 3 and add it to its mean value and then designate the calculated number X, we can obtain an X larger than 7. Because we designed our questionnaire using a 7-point Likert Scale, X values larger than 7 are not usual. We can arrive at the conclusion that the variation in our data is considerable, which could be the cause of the insignificance in the results of our analysis.
(3) Current User Attitudes toward cloud computing i. Excessive Reliance of Public Cloud Users
According to a survey regarding cloud computing conducted by TechSoup Global, a nonprofit organization, a lack of training in a company ranks high among the factors that hinder companies’ adoption of cloud computing. In addition to training issues, the willingness of companies to outsource public clouds to a third party is corroborated by other evidence that indicates that additional expenses and unfamiliarity with cloud computing also drive companies to turn to a third party
N Mean
Standard Deviation
Standard
Error Minimum Maximum
Internet
‧
increasing the access of one’s company to different technological possibilities. A survey conducted by Institute For Information Industry shows that less than half of the corporations in Taiwan own, develop, or maintain their own IT hardware equipment and that only 34.9% of all corporations develop and maintain their own software resources. In addition, as for public cloud users, who use computing and IT infrastructure that is owned by an organization (Harris, 2011), the survey shows that a low percentage (42.3%) of companies are willing to own and maintain their IT infrastructure. Many previous surveys and research indicate that outsourcing is one of the most popular strategies for expanding companies’ technological capabilities, whether in the area of public clouds or in other areas. However, this phenomenon may give rise to other problems. From the perspective of service receivers, outsourcing causes them to have less control over both the quality and the operation of their IT projects because they are managed by people who are beyond their range (Grover and Teng, 1993). In turn, the loss of control increases the reliance of the service receivers on their service providers. In table 5-1, we can observe that all of the numbers in the mean difference column are negative, which suggests that companies using public clouds already lack a focus on issues regarding internet connections, contract management, and IT staff training because they rely on third parties to take care of these issues. Therefore, we reason that the excessive reliance of public cloud users is a discernible phenomenon.ii. User Resistance Effects
The best method of succeeding in the adoption of new technology has long been a contested issue among researchers because the underlying affecting factors can vary depending on the field. Understanding the factors that contribute to the success of implementation is a central concern. One of the key factors from which many implementation problems stem is user resistance. Different characteristics of different types of systems often result in organizational and operational revolution and thus may be resisted by users (Jiang, Muhanna, and Klein, 2000). In our research, we attempt to determine whether the factors internet connections, contract management, and IT staff training are part of readiness for public clouds. The results show that companies that have adopted cloud computing do not have better internet connections, contract management, and IT staff training. The results indicate that because of user resistance, the employees of these companies tend to resist these changes brought about by the new implementation of public clouds and that their resistance leads to dissatisfaction with their company’s operation. For example, when companies must contend with contracting a third party to provide services, it is difficult for those
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
companies to be well-rounded and use a sophisticated method or strategy initially.
Most of the time, they face unsuspected problems even when precautionary measures have been taken. By the same token, companies adopting public clouds are often lacking in good systems for providing staff with proper training, even when they have begun to incorporate public clouds into daily use. Therefore, in Table 5-1, we observe that public cloud companies have smaller mean values than non-cloud-computing companies.
Finding 4: Scatter diagrams of the returned data for companies that have adopted either public clouds or public clouds
We analyze our returned data using SmartPLS and the result includes weights for different public or private cloud readiness factors. Besides, we design our questionnaire using a 7-point Likert Scale to assess different factors. Based on the weights and scale, we generate the equations as follows.
Public clouds:
TIC MCM MIS Y = 0.48X1 + 0.42X2 + 0.1X3 Private clouds:
MIS TV TII MSS Y = 0.04X1 + 0.21X2 + 0.29X3 + 0.46X4
We then use the equations to count the score of current companies’ cloud status.
We show the result of the counting in percentage and we magnify every score proportionally to make the largest score of the returned data into 100%. For example, if the largest score is 86%, we multiply the number by a curtain fold to turn it into 100%, and multiply the rest of the scores by the same certain number to magnify them proportionally. The results are shown as follows.
Table 5-7. Scores for Public Clouds
33.33% 34.27% 40.96% 53.52% 55.99% 61.50% 64.67% 65.26% 66.04% 66.67%
66.67% 67.84% 68.70% 71.13% 73.00% 74.45% 76.52% 77.47% 77.82% 77.86%
78.01% 78.52% 79.50% 82.51% 87.17% 89.32% 91.00% 93.35% 94.52% 99.18%
100.00%
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Table 5-8. Scores for Private Clouds
49.91% 56.19% 56.88% 57.24% 58.29% 59.08% 61.18% 62.67% 64.13% 65.97%
67.80% 69.73% 70.16% 70.50% 71.36% 71.71% 72.96% 73.82% 73.82% 74.79%
74.80% 75.13% 76.27% 79.07% 80.37% 81.35% 85.10% 85.95% 86.90% 89.53%
100.00% 100.00% 100.00%
We then draw two scatter diagrams based on the results to show the current status of cloud adoption. Additionally, the statuses of company are sorted in ascending order. The graphical results are as follows.