Elsevier

Future Generation Computer Systems

Volume 101, December 2019, Pages 502-523
Future Generation Computer Systems

Value-based cloud price modeling for segmented business to business market

https://doi.org/10.1016/j.future.2019.06.013Get rights and content

Highlights

  • Cloud pricing should consider both cloud service providers’ cost and customers’ value propositions.

  • Cloud market segmentation can differentiate various customers’ values.

  • Value-based cloud pricing provides a better solution for a cloud service provider to maximize its profit.

  • Four types of cloud pricing models illustrate a comprehensive framework of the cloud pricing process.

  • Genetic Algorithm offers a convenient solution of optimizing each optimal price for each pricing model.

Abstract

Cloud price modeling is the major challenge facing many cloud computing practitioners and researchers in the field of cloud economics, which is also known as “Cloudonomics.” Previous attempts mainly focused on a uniform market and used existing price models to explain the issue of revenue maximization for cloud service providers (CSPs) from a cost or internal rationality perspective but paid less attention to the cloud market segmentation for cloud business customers from a surplus value or external rationality perspective. This study considers both aspects of the value proposition. Based on the assumptions of the customers’ utility values for different market segments, we establish a framework of value-based pricing strategy and demystify the process of modeling and optimizing cloud prices for CSP to maximize its profits. This framework is built upon the theory of value co-creation for both customers and CSPs to form a business partnership. We show how to create four cloud pricing models, namely: on-demand, bulk-selling, reserved, and bulk + reserved. We also demonstrate how to identify the optimal price point of each model to maximize CSP’s profit by genetic algorithm. We exhibit that reserved, bulk + reserved, on-demand, and bulk-selling can deliver a profit margin of 203%, 183%, 166%, and 157% for CSPs respectively. While the reserved model provides the highest profit margin, it does not necessarily mean that CSPs should adopt one model only. We provide a novel solution that allows CSPs to achieve the maximum profit by offering multiple pricing models simultaneously to various customers in the segmented market. We argue CSPs should capitalize on cloud pricing rather than price to gain more cloud market share and profit. Thus, we present state-of-the-art cloud pricing for segmented business to business cloud market.

Introduction

Value-based cloud price modeling for different cloud market segments [1] is vital to all Cloud Service Providers (CSPs) as it will not only impact on CSP’s profitability but also determine the sustainability of CSP’s cloud business [2]. The goal of this study is to develop a comprehensive process framework of value-based price modeling that enables CSP to gain more cloud B2B market share for its profit maximization. Many previous studies can be considered as either cost-based or cost-plus models [3], which they were dependent on an assumption of a uniform market and paid less attention to the segmented market that carries heterogeneous values of customers. Furthermore, their processes of modeling mainly explained how to leverage two or three existing models (e.g., on-demand and spot instance) for CSP to maximize its revenue, which was subjective to a cloud capacity constraint that is equivalent to a cost. Subsequently, those works led to the issue of the internal rationality only.

The term of “Rational” means a decision is made according to reason or logic. In economics, people are assumed to be rational because they will systematically and purposefully do the best they can do achieve their purposes, given the available choices [4]. “Internal rationality” implies that a decision maker focuses on internal justification; for instance, a cloud price is determined by cost. In contrast, “external rationality” suggests that a decision should be made by an explanation of external factors, which a price is dependent on customer willingness to pay. In economics, it is essential that the pricing model is built upon the assumption that the individual is rational because people can be irrational.

The questions of how to create a cloud price model itself based on the business customers’ value proposition and how to target the segmented market, especially, business to business (B2B) market have remained either unanswered or incomplete. To overcome this gap, we develop cloud price models that include both external and internal rationalities. For the external rationality, we examine two essential external factors, namely, cloud customers’ utility values and B2B market segments. For the internal rationality, we take into account of CSP’s cloud infrastructure cost. Based on our result of price modeling, we then use a genetic algorithm (GA) to identify the optimal price point of each price model for CSP to maximize its profit. One of the useful properties of GA is that it can solve a complex profit equation for intertwined variables without knowing the details of sub-functions. It is also convenient to upgrade the optimal price point of each model so that the process of price modeling can cope with the decision variation of cloud business strategy.

To demonstrate the process of value-based cloud price modeling, we exhibit and analyze different models, namely cost-plus, on-demand, bulk-selling, reserved (two-part tariff), and reserved + bulk for profit margin comparison. The cost-plus pricing models are often prevalent [2] because “they carry an aura of financial prudence… to yield a fair return on overall costs (or resources), fully and fairly allocated”. However, these models fail to capture heterogeneous values of customers. In contrast, four value-based models can reflect the value proposition of both cloud customers and CSPs. Those models can be considered as “value co-creation” [5], [6] because CSPs are seeking a partnership with their cloud customers in the cloud market value chain. We show these models allow CSPs not only to satisfy customers’ needs but also to achieve a better profit margin in comparison with the cost-based model. Overall, we provide a process solution that has a quantitative measurement under a single currency (or business revenue contribution) can capture different cloud customer service metrics (e.g., increase sales, customer retention, investment efficiency, maintain a specified SLA, reduce checkout queueing time, etc.). To better illustrate the entire process of modeling, we use the following scenario to explain the details.

Assume a group of decision-makers of a hosting firm decide to expand its hosting business into the cloud B2B market. It implies that the firm wants to become a new CSP to compete with other existing CSPs (either global or local CSPs). If the initial investment budget (both capital and operation expenditure or Capex and Opex) and business goals (targeted revenue, profit, and market) have been approximately identified, the decision makers want to know how to achieve the business goals. There are two fundamental questions must be clarified: “How does the firm form the right pricing strategy for the identified business goal?” and “how does it decide the appreciated cloud price models along with optimal price points, sales volumes, and unit cost to achieve the maximum profit?” These questions will help the CSP to divert its limited resources (investment budget and technology expertise) to serve its targeted customers better so that it can maintain its cloud business profitability and sustainability. There are many possible pricing strategies to reach the business goal, namely cost-based, market-based, and value-based pricing. As Hinterhuber [7] indicated, both cost-based (37%) and market-based pricing (44%) are much popular than value-based pricing (17%). Nagle [2] observed that historically, cost-based pricing is the most common pricing strategy in most industries because “in theory, it is a simple guide to profitability; in practice, it is a blueprint for mediocre financial performance”. Unfortunately, the issue of the cost-based pricing strategy is when there is strong market demand, the average unit cost will decline, and the price reduction should follow because the profit margin is determined by the unit cost (e.g., 30%–100%). Conversely, when the market demand becomes weak, the average unit cost will go up, and the price should be raised. It contradicts a sensible pricing strategy in term of market response.

The alternative way of cost-based pricing is either market-based (or competition-based) or value-based (customer-driven) pricing. Market-based pricing is to set cloud service price based on the current competition condition or equilibrium of supply and demand. However, competition-based pricing could mislead CSPs to see market-based pricing as a zero-sum game, which what the customers’ gain is the CSP’s loss [8]. They might also believe they do not influence price because market-based pricing is a competition behavior of the market. In contrast, value-based pricing can offer customer needs and create real value to satisfy customers because it is determined by customers’ utility values.

Nonetheless, the definition of value-based pricing can be subject to a wide range of interpretations. It is dependent on the context of the content. The term is often defined as a pricing process for an individual’s preference (ordinal utility [9]) that aims to the B2C market. However, this paper of value-based pricing focuses on the marginal value (cardinal utility) that aims to the B2B market. It implies the process is to capture a proportional of value that CSP might impact on the targeted cloud customers for their business [8]. In other words, a CSP is to develop and deliver the cloud service values for its cloud customers to achieve business success and then seek a reward for its distributed services.

In general, the B2B market emphasizes the entire value chain and partnership development. The purchasing decision is not made by single or few individuals, but by more than dozens of stakeholders for the cloud service values that CSP can offer. Therefore, value-based pricing becomes one of the effective pricing strategies for the cloud B2B market. The cloud services can influence the customer’s business in term of increasing their profit margin, higher business revenue and lower the operation cost.

Overall, the framework of value-based pricing strategy includes (1) identifying target customers and workload patterns that are related to each cloud market segment, (2) quantifying cloud customer utility functions that are associated with service values and cloud service metrics, (3) establishing various cloud pricing models based on the specified customer’s utility functions, (4) identifying the optimal price points for CSP to achieve the total maximization profit from all market segments. Fig. 1 presents this processing framework of price modeling of all elements.

Due to the limited space, it is impossible to include all four elements of a value-based pricing strategy into a single paper. This study will only focus on developing cloud pricing models (element 3) and identifying the optimal price points (element 4). The other elements of the value-based strategy have been presented in separated papers, and one of them has been published [10]. However, we will lay out enough details of cloud market segmentation (element 1) and customer utility functions (element 2) [11].

The purpose of cloud market segmentation is to gather cloud customers’ usage patterns so that a CSP can work out a right pricing strategy to serve its targeted customers well while it can achieve its maximization profit within its budget constraints. In fact, Yankelovich [12] specified the detail criteria of market segmentation: (1) align with the company’s strategy, (2) specify where the revenue and profit come from, (3) articulate customer’s business values, (4) focus on actual business behaviors, (5) make sense to the firm’s executive team and the board and (6) to be flexible to accommodate or anticipate market changes quickly. According to these market segmentation criteria, we develop a novel solution [10] that allows CSP to identify the cloud B2B market segment quickly. The solution is a combination of hierarchical clustering (HC) with time-series (TS) methods according to two datasets, which one is from Google public dataset [13] and other is extracted from a local hosting firm for its hosting business. From Google’s dataset, we can develop six potential cloud market segments that are determined by the number of parameters of usage patterns, such as job priority, cores quantity, memory size, and AMD’s virtualization workload guidelines [14]. This number of cloud market segments is within the range of McDonald’s suggestion [15], which the suggested number of the market segment is between 5 and 10.

The results of cloud market segmentation are shown as in both Table 1 and Fig. 2. The details discussion of these market segments associated with utility function has been presented our previous publications [10], [11]. Once the cloud market segments have been quantified, the next issue is how to develop the cloud customers’ utility functions for these cloud market segments.

The goal of modeling cloud customers’ utility function is to quantify the cloud customers’ experiences and preferences (utility values) that are subject to the cloud resources provision. Practically, these subjective experiences concern the running applications for cloud customers to generate business revenue or profit, which can be quantified by the service metrics [19].

The meaning of utility is quite ambiguous because it consists of different connotations. Historically, the implication of utility was derived from utilitarianism. It means a subjective experience and satisfaction. It is known as the utilitarian tradition. Later, this term has been extended to the contractarian tradition, which emphasized social welfare [20]. As a result, the contemporary meaning of utility has three connotations:

  • (1)

    The economic utility refers to subjective satisfaction and happiness. “It is an alternative way to describe preference and optimization” [4] The utility value in this context is measured by different preferences under information uncertainty in term of risks and wealth.

  • (2)

    Another implication of utility is an essential infrastructure service for the public. Sometimes, it is also called as “public utility”, such as water, electricity, and telephone service that are supported by some incumbent providers. It is associated with the term of social welfare.

  • (3)

    “Utility” also refers to the utilization rate. It is measured by a percentage value between 0 and 1. For example, the utility of a network means its utilization rate. It is a concept of efficiency. It is different from the economic connotation of utility that is measured by preferences.

However, there are many research works that assume both economic utility and utilization rate are the same. The utilization rate can be included in a cloud service metric, but it is not the same as the utility value in an economic sense. Economically, a business customer’s utility represents the amount of business revenue or profit that is contributed by the number of VMs (e.g., wealth). For instance, the utility of a mission-critical application will be totally different with the utility of backend type of workloads, such as log data processing or MapReduce [19] because the end users will pay a different price for the cloud services. The question is, how we can use a single currency to reflect various utility values and align with CSP’s profitability? To solve this issue is to unify all customers’ utility values and CSP’s profit into a measurement of cloud customers’ business revenue or profit. This is also known as value co-creation. The benefits of value co-creation are that CSPs can reduce investment risk and maintain cloud customer loyalty [21] and uphold CSP’s profitability and business sustainability. The modeling process of quantifying customers’ utilities is to establish a relationship between the customer’s business revenue or profit contribution (a dependent variable) and the number of VMs (independent variable) to be provisioned.

Based on different characteristics [22] of the cloud business applications, we organize utility functions into three categories:

  • Utility functions (Segment 4 and 5) are defined by High Availability (HA) characteristics [23], [24], [25].

  • Utility functions (Segment 1 and 3) are determined by response time characteristics [26].

  • Utility functions (Segment 2 and 6) are identified by risk characteristics (risk-averse, risk-seeking, and risk-neutral [19]).

The process of how to quantify these utility functions is presented in the paper [11]. Table 2 highlights the result of six utility functions. (Details assumptions of these functions are presented in Section 3.2.1.) Now, the subsequent questions are how we can build various price models for a CSP to capture more cloud market share and how to identify the optimal price point of each model for profit maximization? These problems are what we will focus on in this research.

By microeconomics [27], we can formalize the CSP’s profit problem into the following equations. Eqs. (1), (3) mean the total business profit is dependent on a sales price, an average unit cost (or a marginal cost), and sales quantity (e.g., market demand). Intricately, the quantity is a function of a price, and the price is an inverse function of the quantity. Mathematically, we can present this interdependent relationship in Eq. (2) πp=RpCQp CQp=cuQpQp,Rp=pQp,p=Q1p πp=QppcuQp=pQpCQpwhere πp is a cloud business profit, Rp is a cloud revenue, CQp is the total cost, p is a unit price and cuQp is the average unit (or marginal cost) which is also a function of the total sales quantity Qp.

The issue is how we can achieve the maximum profit by identifying the optimal price point Eq. (4). While the equation appears evident and straightforward, it is difficult to find a clear solution because of both functions Q(p) and p=Q1p are generally unknown p=argmaxpπp

The primary challenge is that the relationship of p=Q1(p), cuQp, and Qp is intertwined. Moreover, these equations will become progressively more complex if various pricing models are introduced.

Previous works solve the problem by excluding the cost component from a profit Eq. (1) [28] or by making some restricted assumptions [29] [30], or by assuming a uniform market that is derived from α-fair utility [31]. Others assume a price is a simple linear equation based on the AWS’ historical data within a coefficient band [32]. Still, others intend to offer a solution by mixing with existing on-demand, reserved and spot instances from a CSP’s perspective [29]. Although their works have made excellent progress in the context of cloud price modeling for the B2C market, many critical aspects of modeling remain unanswered. This study provides various solutions to resolve these issues. These solutions encapsulate the comprehensive process framework of value-based pricing strategy.

By providing the various solutions, this work has made the number of contributions:

  • To the best of our knowledge, it is the first time to create various cloud price models based on market segmentation theory and the number of utility functions that are defined by cloud customer business revenue or profit contribution.

  • This work has clearly illustrated how to establish four value-based price models according to the defined business strategy

  • By leveraging the actual retail pricing experiences, this study develops bulk-selling and reserved models for CSP to have more pricing options to achieve a higher profit margin.

  • This work also illustrates the relationship between bulk-selling and bundle services. By developing various cloud pricing models, CSPs can spontaneously launch more pricing models to capture more profit across various market segments.

  • We demonstrate how to apply GA to identify the optimal price point for each price model.

  • The price models are dependent on both internal (CSP’s cloud infrastructure costs) and external (cloud market segments and customer utilities) rationality.

  • This paper presents novel and practical solutions so that many practitioners can plug in their datasets and build their own price models based on the defined company’s business strategy.

  • Most importantly, this study shows how to calculate the total revenue and profit based on different pricing models that are offered to various customers spontaneously.

The rest of the paper is organized as follows: Section 2 provides a brief overview of related works in cloud pricing. Section 3 formalizes four value-based pricing models according to various assumptions with different constraints. Section 4 presents concise information about genetic algorithms (GA) and how to determine the GA parameters for our experiments. Section 5 shows the experimental results. Section 6 offers a detailed analysis of cloud pricing and optimization. Section 7 concludes the paper and proposes future research directions.

Section snippets

Related work

In light of the value theory [33], we can approximately classify most of cloud price models into three basic categories, namely value-based, market-based, and cost-based pricing. The value-based pricing is often considered as a subjective view of the cloud pricing from a demand side because it concerns the measurement of customers’ subjective experience and utility preferences. The cost-based pricing is regarded as an objective view of cloud pricing from a supply side because it is built on the

Market assumptions

According to the theory of the B2B market [8], the cloud B2B market is a relational business market because it emphasizes building a mutual value-creation relationship or partnership with business customers. It requires long-term relationship development. In contrast, business to consumer (B2C) market mainly is focusing on the final transaction between a firm and an end-user [8]. From this perspective, we will first consider the cloud price models based on the assumption of a monopoly market 

Proposed methods

There are many possible optimization methodologies or techniques that we could apply for the optimizing problem, such as gradient descent, Genetic Algorithm (GA), and simulated annealing. Gradient descent cannot be applied because the profit equation is noncontiguous. Simulated Annealing could be one of the possible methods for our problem because it usually is better than greedy algorithms, but the technique can be slow, especially if the cost function is expensive to compute. Subsequently, we

On-demand pricing model results

Table 7 shows the final results for all pricing models that are including on-demand, which CSPs should charge $0.749 per VM instance/hour for the maximum profit of $2,463. The average unit cost is about $0.281. In comparison with cost-based pricing, the on-demand pricing can boost a 66% profit margin if we take account of the external rationality. Although the profit margin (100%) of the cost-based pricing looks very attractive, it is not optimal. The result of this comparison means the

Analysis and discussion

This study demonstrates a comprehensive framework of how to formulate four value-based cloud pricing models from a customer’s value co-creation perspective. In contrast to previous works that assumed a uniform market with only one utility function, this solution of cloud pricing is much realistic and practical because market segmentation practice has been widely applied to many service industries. Cloud industry is not exceptional. AWS has adopted up to seven different types of pricing models

Conclusions and future work

This study has developed an overall framework of the pricing process that is how to generate various price models and how to find these optimal price points of each model for CSP to maximize the profit. These are two elements of pricing strategy (Shown in Fig. 1) that have been demystified in our research work. The significance of this study is that it presents a comprehensive and practical process for value-based pricing.

We demonstrate how to establish four types of practices price models,

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 and ACM member and has just completed his Ph.D. journey at the University of Melbourne recently. He is the first author of Cloud Data Center and Cost Modeling. He was a senior IT and network design engineer in Telstra for nearly 20 years. He designed, built 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.

References (72)

  • ClaycampH. et al.

    A theory of market segmentation

    J. Mark. Res. (JMR)

    (1968)
  • NagleT.T. et al.

    The Strategy and Tactics of Pricing: A Guide To Growing more Profitably

    (2011)
  • LeardiR.

    Nature Inspired Methods in Chemometrics: Genetic Algorithms and Artificial Neural Networks

    (2003)
  • MankiwG.N.

    Principles of economics

    Cengage Learn.

    (2014)
  • RamaswamyV. et al.

    What is co-creation? An interactional creation framework and its implications for value creation

    J. Bus. Res.

    (2017)
  • KohtamakiM. et al.

    Theory, and practice of value co-creation in B2B systems

    Ind. Mark. Manage.

    (2016)
  • HinterhuberA.

    Customer value-based pricing strategies: Why companies resist

    J. Bus. Strategy

    (2008)
  • GrewalR. et al.

    Handbook of Business-To-Business Marketing

    (2012)
  • IkefujiM. et al.

    Pareto utility

    Theory Decis.

    (2013)
  • C. Wu, R. Buyya, K. Ramamohanarao, Cloud computing market segmentation, in: Proceedings of the 13th International...
  • WuC. et al.

    Modeling cloud business customers’ utility functions

    (2019)
  • YankelovichD. et al.

    Rediscovering market segmentation

    Harv. Bus. Rev.

    (2006)
  • https://github.com/google/cluster-data/blob/master/TraceVersion1.md (Accessed 20 August...
  • https://www.amd.com/Documents/AMD_WP_Virtualizing_Server_Workloads-PID.pdf (Accessed 20 August...
  • McDonaldM. et al.

    Market Segmentation How to Do It and How to Profit from It

    (2012)
  • MichalskiO. et al.

    Implementing Azure Cloud Design Patterns

    (2018)
  • WuC. et al.

    Cloud Data Center Cost Modeling, a Complete Guide To Planning, Designing and Building a Cloud Data Center

    (2015)
  • TarasovV. et al.

    Virtual machine workloads: the Case for new benchmarks for NAS

    Fast

    (2013)
  • MishraA.K. et al.

    Towards characterizing cloud backend workloads: Insights from google compute clusters’

    Perform. Eval. Rev.

    (2010)
  • JehleG.A. et al.

    Advanced Microeconomic Theory

    (2011)
  • De KoningJ.

    Service design geographies

  • CalzarossaM.C. et al.

    Workload characterization: A survey revisited

    ACM Comput. Surv.

    (2016)
  • A. Undheim, P. Heegaard, Differentiated availability in cloud computing SLAs, in: 2011 IEEE/ACM 12th International...
  • ZhangM.

    Utility function in autonomic workload management for DBMSs

    Int. J. Adv. Intell. Syst.

    (2012)
  • MaurerM. et al.

    Cost–benefit analysis of an SLA mapping approach for defining standardized cloud computing goods

    Future Gener. Comput. Syst.

    (2012)
  • WilliamsA. et al.

    Web Workload Characterization: Ten Years Later

    (2005)
  • BelleflammeP. et al.

    Industrial organization markets and strategies

    (2011)
  • ToosiA.N. et al.

    Revenue maximization with optimal Capacity control in infrastructure as a service cloud markets

    IEEE Trans. Cloud Comput.

    (2015)
  • XuH. et al.

    Dynamic cloud pricing for revenue maximization

    IEEE Trans. Cloud Comput.

    (2013)
  • AazamM. et al.

    Cloud customer’s historical record based resource pricing

    IEEE Trans. Parallel Distrib. Syst.

    (2016)
  • XuH. et al.

    A study of pricing for cloud resources

    ACM Sigmetrics Perform. Eval. Rev.

    (2013)
  • Agmon Ben-YehudaO. et al.

    Deconstructing amazon EC2 spot instance pricing

    ACM Trans. Econ. Comput.

    (2013)
  • HiroseI. et al.

    The Oxford Handbook of Value Theory

    (2015)
  • AltmannJ. et al.

    Cost model-based service placement in federated hybrid clouds

    Future Gener. Comput. Syst.

    (2014)
  • M. Macías, J. Guitart, A genetic model for pricing in cloud computing markets’, in: Proceedings of the 2011 ACM...
  • NovaniS.
  • Cited by (0)

    Caesar Wu is a senior IEEE and ACM member and has just completed his Ph.D. journey at the University of Melbourne recently. He is the first author of Cloud Data Center and Cost Modeling. He was a senior IT and network design engineer in Telstra for nearly 20 years. He designed, built 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.

    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, and four textbooks are recognized as a “Web of Science Highly Cited Researcher” both in 2016 and 2017 by Thomson Reuters, a Fellow of IEEE, and Scopus Researcher of the Year 2017 with Excellence in Innovative Research Award by Elsevier for his outstanding contributions to Cloud computing.

    Ramamohanarao(Rao) Kotagiri received Ph.D. from Monash University in 1980. He has been at Melbourne University since 1980 and was appointed as a professor in 1989. Rao was Head of CS and SE and Head of the School of EE and CS. He received Distinguished Contribution/Service Awards from ICDM, PAKDD, DASFAA, etc. Rao is a Fellow of the Institute of Engineers Australia, Fellow of Australian Academy Technological Sciences and Engineering and Fellow of Australian Academy of Science.

    View full text