Modeling cloud business customers’ utility functions
Introduction
The goal of this study is to define multiple utility functions for different cloud business customers’ preferences that are grouped into various cloud market segments so that a Cloud Service Provider (CSP) can create a price strategy based on a broad spectrum of cloud market to maximize its profit. Moreover, the CSP can tailor its limited investment resources and technical expertise to serve its target customers effectively.
In economics, the concept of utility means measuring a choice of individual’s preferences. In other words, it is to evaluate the individual’s subjective satisfaction, happiness and perception of worthiness that “the consumer derives from consumption of goods and services”. [1] The subjective measurement of the utility value reflects on an acceptable price that the individual is willing to pay for [2]. This acceptable price leads to an idea of the utility function definition, in which a subjective value is dependent on the number of goods or services (e.g., Virtual Machine or VM) to be consumed or provisioned. According to Krugman and Wells [1], different individuals would have different utility functions because different people would have different needs and preferences towards a certain amount of goods or services.
However, this economic term of “utility” is often mixed with other connotations of “utility” so that it becomes quite ambiguous and confusing [3] when a utility function is defined. It is necessary to clarify and differentiate meanings of utility at the outset.
The common sense of utility means “the usefulness of something, especially in a practical way”. For example, the utility of database means to implement various processes or functions of the database, such as batch update, rebuild, recovery, backup, etc. Another sense of utility is quite close to the meaning of the usability that often refers to the state of being useful, which is to supply the essential infrastructure services to the general public. These services are offered by incumbent service providers, which are known as “public utilities” or simply, “utilities”. For example, Buyya et al. [4] argued “cloud computing” is the 5th utility. Still, another meaning of utility is the utilization rate, which is to measure the effective usability of something, such as the network’s utility. Its value is between 0 and 1. Although both network and economic utility may adopt a similar function (e.g., isoelastic and alpha-fair function), the contents of two utilities are totally different.
Economically, the utility function is to describe how people consume various amounts of goods and services in terms of their subjective preferences, needs, and experiences in a less or more rational way. “It is simply a convenient device for summarizing the information contained in the consumer’s preference relation” [5]. This preference is measured by either cardinal or ordinal approaches. “Cardinal” means a marginal value can be quantified by an additional subjective value for one more unit of cloud resources that are acquired. The ordinal approach can only be measured by a ranking method. Our study will adopt a cardinal approach [3] to quantify the cloud utility values because the cloud utility satisfies the criteria of cardinal analysis: (1) the cloud business customers are rational. It means they will systematically and purposefully do the best they can do to achieve their goals, given the available choices, (2) Utility value can be measured numerically in terms of dollar value, and (3) Unit of Infrastructure as a Service (IaaS) is homogeneous. Subsequently, we can define various utility functions in the cloud context.
If a CSP focuses on the business customers, we can consider the customers’ utility is equivalent to a cloud customer’s business revenue or business income so that a utility function can be defined as a function of the independent variable: “q” (cloud resources or a quantity of VM). Therefore, we can define it as , where is the number of cloud market segments. This is determined by a market segment assumption and CSP’s business and marketing strategy [6].
The focal point of this paper is to create different types of utility functions for various cloud business applications, such as web hosting, content delivery, e-commerce (e.g., an online check out system), database backup, disaster recovery (DR), virtual desktop infrastructure (VDI), and backend processing (e.g., MapReduce, log file analysis). If we assume that the measurement of the business customer’s satisfaction (e.g., cloud service metrics) is directly associated with its business revenues, then our modeling process is to estimate how much the customers are willing to pay for a given quantity of the cloud resources that can help them to grow their business revenue or profit. Fig. 1 ( means a profit) provides an overall cloud pricing strategy. There are four necessary steps in the process. The 2nd step highlights the focal point of this paper, which is to define utility functions for CSP to achieve the value co-creation with its business customers or partners.
Fig. 1 provides details on how we create a cloud pricing strategy. The 1st step can be found in our early work of cloud market segmentation [6]. The 3rd step is how to establish multiple cloud pricing models to meet various customers’ requirements [7]. The 4th step is to identify the optimal price point of a pricing model for a CSP to achieve profit maximization [7].
According to Nagle et al. [8], the 2nd step is a challenging task because the issue requires multidisciplinary knowledge. Many previous types of research either ignored some parts of the problem or failed to articulate the problem clearly. Consequently, the issue of defining a utility function becomes hard to be implemented in practice. Many previous solutions of utility modeling [9], [10], [11], [12], [13], [14], [15], [16] often assume a uniform market or “one size fits all”. Moreover, the meaning of utility is often mixed with the demand side of the price and supply side of cost.
To overcome these challenges, this study will clarify the meaning of the cloud customers’ or demand side’s utility for the cloud services, which is a subjective value measurement for the number of VMs to be provisioned. Often, this value can be represented by cloud customer’s experiences (CX) or key performance indicators (KPI) or cloud service metrics (CSM). Both the National Institute of Standards and Technology (NIST) [17] and Oracle [18] have defined CX, KPI, and CSM along with three actual business dimensions that consist of acquisition (increase in sales), retention (monetize relationships), and efficiency (leverage investments). All of the quantitative measurements of business dimensions can be translated into the value of business revenue or profit.
Overall, the core problem of this study is “how to quantify the cloud customer’s satisfaction (the business revenue or profit) along with a variation of “” in each market segment” By defining multiple utility functions related to the segmented cloud market, we provide a novel solution for cloud price modeling that is much more realistic and practicable.
In comparison with previous solutions, such as empirically calibrated price [19] and capacity [20], resource optimization [12], response time, capacity-aware [21], utility-based-SLA [22], and model-based [9], our solution has a number of advantages:
- (1)
It is practical and quantifiable for real cloud applications in term of resource needs (internal rationality),
- (2)
It can be implemented by any CSP for its targeted market,
- (3)
The utility is derived from the principle of economics
- (4)
It is agile and flexible to cope with a CSP’s business strategy changes,
- (5)
The utility functions are defined to improve the cloud customer’s revenue (or external rationality).
- (6)
It is a process of value co-creation for both CSP and cloud customers.
- (7)
It provides a solution for CSPs to achieve more profits by optimizing different cloud price models.
Based on the listed advantages, we have made the following contributions to cloud economics.
(1) To the best of our knowledge, this is the first such study to define multiple utility functions based on the segmented cloud market assumption. It models the utility functions from the value co-creation perspective.
(2) The utility value directly impacts the business revenue or profit of cloud business customers because the utility values are defined by the cloud customers’ KPI or CX or CSM (value proposition). All the utility values can be translated into a single and quantifiable unit rather than some indirect or multiple unquantifiable units. The cloud customer’s utility (or subjective) values become the function of the cloud resources or VM.
(3) We use Markov chain analysis to quantify the number of VM needs in terms of the specified SLA for the high availability (HA) applications, such as database backup, CRM, and terminal server. It provides the deliverable SLA and uptime for customers of segments 4 and 5 (Refer to Table 3) to generate their business revenue or profits. It translates SLA into customers’ business revenue rather than to use SLA as a direct independent variable for the utility function definition.
(4) By leveraging the queueing theory, we create customers’ utility functions that can minimize the response time to deliver the right performance of business applications, such as online checkout system and web-hosting for Segment 1 and 3. This response time length is also translated into customers’ revenue contributions.
(5) In addition, we leverage the concept of risk-aversion and risk-taking to model the customers’ utility values for content delivery (Segment 2) and log file analysis or MapReduce applications (Segment 6) to maximize the end-users’ satisfaction in terms of maximizing application functionality and minimizing cloud running costs.
Overall, our solution can be easily comprehended for CSP decision-makers when they want to make a critical investment decision for their cloud business. Fig. 2 highlights the details of the process of building utility functions. This process illustrates how to define six utility functions by three modeling tools. In summary, Fig. 1 gives an overall cloud pricing strategy, and Fig. 2 shows a novel approach of defining various customers’ utility functions in detail.
The rest of the paper is organized as follows: Section 2 gives a brief literature review regarding previous modeling solutions for utility functions. Section 3 presents how we define multiple utility functions based on the result of six market segments and how we make various assumptions, define the problem, and determine the scaling coefficient and other parameters. Section 4 provides a detailed performance evaluation and validation of our modeling method. Section 5 offers some simple guidelines for CSPs on how to select these utility functions. Section 6 provides both analysis and justification for some assumptions. Section 7 outlines our conclusions and future work. To navigate the discussion easily, Table 1 lists all acronyms used in this paper.
Section snippets
Related work
The modeling customer utility functions for various hosting services or applications can be traced back to the beginning of the dotcom-booming era. Doyle et al. [9], [23] proposed a model-based approach to optimize the hosting of hardware resources for the specified SLA. The goal of their work is to demonstrate how to provision the server resources for web hosting applications effectively. Although the paper adopted the term “utility” and made good progress in hosting service modeling, the real
Prior studies and background
To illustrate our modeling process clearly (Shown in Fig. 2), we can consider a case of how a hosting firm to develop its cloud pricing strategy for its new cloud business. Suppose decision-makers of the firm (supply side) decides to expand its traditional hosting business to a cloud market for its business customers (demand side). The goal of the firm is to grow both revenue (market) and profit with a fixed amount of investment budget.
We also assume that the firm understood its own technical
Performance evaluation
The performance evaluation is divided into two parts. The first part is to compare the market share between our solution of six market segments and other solutions with the single market assumption. The second part is to compare all economic values, which include business revenue, profit, the optimal price, and a unit cost based on the popular price model, namely “on-demand”.
Discussion of model selection (simple guidelines)
Based on various parameters of six cloud market segments, as shown in Table 3, the type of business application can be estimated, which is mapping to each corresponding cloud market segment (See Fig. 3). If an analyst has the real cloud operational dataset, this step will become much more manageable. The crucial issue is how to define the utility function for different customers’ business applications. The basic guidelines can be summarized as follows:
- 1.
If the business customers host a web site
Assumptions analysis
Throughout this paper, we adopt an analytic approach to define multiple utility functions based on various assumptions of business strategy, technology expertise, investment capital, and segmented cloud market so that CSP can explore a broader addressable cloud market to define business customers’ utility functions. If some assumptions are not held or become uncertainty, we can combine analytic, simulation, and statistical approaches. The pre-condition of simulation and statistical method is
Conclusions and future work
The issue of how to define the cloud customers’ utility functions from the cloud customer’s perspective is vital to any CSP because it would help the CSP to generate the optimal cloud price to maximize the profits for its cloud business. Based on our intensive literature review on this topic, we show that one way to improve CSP’s profit is to determine the cloud market segments and then define multiple utility functions from a value co-creation perspective.
Our solution provides external
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Caesar Wu is a senior IEEE member. He is one of the authors of Cloud Data Center and Cost Modeling. He was a senior domain specialist in Telstra. He managed and operated many of Telstra’s enterprises and IT data centers. He has over 30 years working, researching, and academic experiences across various industries. He was the program chair of Computer Information System (CIS) Doctorial Colloquium of The University of Melbourne.
References (64)
- et al.
Cloud computing and emerging IT Platforms: Vision, hype, and reality for delivering computing as the 5th utility
Future Gener. Comput. Syst.
(2009) - et al.
Random utility models in marketing research: a survey
J. Bus. Res.
(2001) - et al.
SLA-based virtual machine management for heterogeneous workloads in a cloud datacenter
J. Netw. Comput. Appl.
(2014) Measuring risk-aversion: The challenge
Measurement
(2016)- et al.
Value co-creation practices and capabilities: Sustained purposeful engagement across B2B systems
Ind. Mark. Manage.
(2016) - et al.
Value co-creation, concept and contexts of application and study
J. Bus. Res.
(2016) - et al.
Cloud pricing models: Taxonomy, survey, and interdisciplinary challenges
ACM Comput. Surv.
(2019) - et al.
Economics Fourth Edition
(2015) Price Theory And Applications
(1995)Encyclopedia of mathematics and its applications. Volume 7. Measurement theory with applications to decision-making, utility and the social sciences Fred S. Roberts
Current Sci.
(2009)
Reny Advanced Microeconomic Theory
Cloud computing market segmentation
Value-based cloud price modeling for segment business to business market
Future Gener. Comput. Syst.
The strategy and tactics of pricing
Oceano-SLA based management of a computing utility
Utility functions in autonomic systems
Resource allocation for autonomic data centers using analytic performance models
Achieving self-management via utility functions
IEEE Internet Comput.
Exploring consumer preferences in cloud archiving – a student’s perspective
Behav. Inform. Technol.
Utility-based resource allocation for virtual machines in cloud computing
Shares and utilities based power consolidation in virtualized server environments
Economic models for cloud service markets: Pricing and capacity planning
Capacity-aware utility function for SLA negotiation of cloud services
Optimizing user’s utility from cloud computing services in a networked environment
Int. J. Adv. Comput. Sci. Appl.
Workflow scheduling algorithms for grid computing
Design of an adaptive framework for utility-based optimization of scientific applications in the cloud
A theory of market segmentation
J. Mark. Res.
Cited by (12)
3D-CUBE readiness model for industry 4.0: technological, organizational, and process maturity enablers
2022, Production and Manufacturing ResearchAn Optimized Cloud-Based Solution for Small-Scale Non-IT Business Firms: An Experimental Analysis
2022, Smart Innovation, Systems and TechnologiesSecurity Threats, Defense Mechanisms, Challenges, and Future Directions in Cloud Computing
2022, Archives of Computational Methods in EngineeringA hybrid multi-criteria decision making algorithm for cloud service selection
2021, International Journal of Information Technology (Singapore)A Study of Costs and Benefits of Content Sharing in Personal Cloud Storage
2021, Journal of Network and Systems Management
Caesar Wu is a senior IEEE member. He is one of the authors of Cloud Data Center and Cost Modeling. He was a senior domain specialist in Telstra. He managed and operated many of Telstra’s enterprises and IT data centers. He has over 30 years working, researching, and academic experiences across various industries. He was the program chair of Computer Information System (CIS) Doctorial Colloquium of The University of Melbourne.
Dr. Rajkumar Buyya is a Redmond Barry Distinguished Professor and Director of the CLOUDS Laboratory at the University of Melbourne. He has authored more than 600 publications. He is recognized as a ”Web of Science Highly Cited Researcher” both in 2016 and 2017 by Thomson Reuters, a Fellow of IEEE, and Excellence in Innovative Research Award by Elsevier for his outstanding contributions to Cloud computing.
Dr. Kotagiri Ramamohanarao (Rao) is received the Ph.D. degree from Monash University. He was awarded the Alexander von Humboldt Fellowship in 1983. He was Research Director for the Cooperative Research Center for Intelligent Decision Systems, the program co-chair for VLDB, PAKDD, DASFAA, and DOOD conferences. He was awarded Distinguished Contribution Award in 2009 by the Computing Research and Education Association of Australasia. He is a Fellow of Australian Academy of Science, a Fellow of Australian Academy of Technology and Engineering and a Fellow of Institution of Engineers Australia.