Two-stage game theoretical framework for IaaS market share dynamics

https://doi.org/10.1016/j.future.2019.08.003Get rights and content

Highlights

  • We propose a Stackelberg game to capture the user demand preferences.

  • A differential game is proposed for IaaS providers to compete over service quality.

  • Two-stage game allows the new IaaS Providers to have a share in the market.

  • Experiments proved the best strategy is to increase service cost and quality.

Abstract

In this paper, we consider the problem of cloud market share among Infrastructure as a Service (IaaS) providers in a competitive setting. The public cloud market is dominated by few large providers, which prevents a healthy competition that would benefit the end-users. We argue that to make the cloud market more competitive, new providers, even small ones, should be able to inter this market and find a share. This problem of deeply analyzing the cloud market and providing new players with mechanisms allowing them to have a market share has not been addressed yet. In fact, to make the cloud market open and increase the cloud service demand, we show in this paper that the cloud providers have to compete not only over price, but also quality. Most of the research performed in the cloud market competition focus only on pricing mechanisms, neglecting thus the cloud service quality and user’s satisfaction. However, to be consistent with the new era of cloud computing, Cloud 2.0, providers have to focus on providing value to businesses and offer higher quality services. As a solution to the aforementioned problem, we propose a conceptual, user-centric game theoretical framework that includes a two-stage game: 1) to capture the user demand preferences (optimal capacity and price), a Stackelberg game is used where IaaS providers are leaders and IaaS users are followers; and 2) to enhance the service ratings given by users in order to improve the provider position in the market and increase the future users’ demand, a differential game is proposed, which allows IaaS providers to compete over service quality (e.g., QoS, scalability and adding extra features). The proposed two-stage game model allows the new IaaS providers, even if they are small, to have a share in the market and increase user’s satisfaction through providing high quality and added-value services. To validate the theoretical analysis, experimental results are conducted using a real-world cloud service quality feedback, collected by the CloudArmor project. This research reveals that due to the fact that service customization tends to enhance the customers loyalty in today’s subscription cloud economy, the best strategy for small IaaS providers is to increase the service cost and improve the quality of their added-value solutions to prevent customers’ defection. This not only elevates the provider’s profit, but also increases the quality equilibrium that leads to a higher user satisfaction. Consequently, higher satisfaction enhances the provider’s rating and future users demand.

Introduction

The rising demand in the cloud infrastructure service market has tempted a large number of technology providers to participate and compete in the market [1]. However, today’s cloud market is dominated by only few large providers. As reported by the Synergy Research Group 2017,1 Amazon, Microsoft, Google, and IBM gained ground in the market at the expense of smaller IaaS providers. The medium sized IaaS providers lost 1% of the market and a large number of small IaaS providers collective market share dropped by 4%, as illustrated in Fig. 1. Such a dominated market prevents a healthy competition. It also hinders compatibility with private clouds and prevents offering personalized added-value services by resellers [2]. Lack of these services may threaten the wide adoption of cloud computing in many industries. Thus, for the growth of the cloud computing industry, there is an increasing need to open the market to the new and smaller providers and create a more competitive environment.

Cloud IaaS has been in the center of attention for years and several research proposals about the technology itself have been lunched. Nonetheless, there is an urgent need to explore and address the business issues surrounding cloud computing, while considering the technical characteristics of such a paradigm. Nowadays, online market and rating platforms made it easy for users to compare a wide range of infrastructure services and for IaaS providers to establish their own credibility. In this paper, we argue that each IaaS provider entering the market needs to distinguish itself from the already established players and compete over both price and quality. As outlined in [3], today’s market of Cloud 1.0 is price-focused. For that reason, there are extensive research that considered pricing competition and proposed optimal pricing strategies in order to maximize the final revenue of cloud providers [4], [5]. However, there is a large number of modern business applications for which a price-focused service model will not be adequate. Often, users hesitate to move their critical business process to the cloud since the first-generation cloud obscured its operations detail behind its low pricing models [3]. Hiding the details blurs the vision of customers about the trade-offs that the IaaS provider has made in order to offer computing at such a low price.

The new era of cloud computing, Cloud 2.0, has been emerged to focus on providing value to small and medium enterprises (SME) as well as large enterprise markets at higher costs as well as higher quality [3], [6], [7], [8], [9]. For the revolution of Cloud 2.0 to take place for IaaS, two transformations need to occur: (1) IaaS providers must be prepared to provide value to businesses that entices them out of their built-in IT resources and applications; and (2) customers must demand a combination of fast, secure, and reliable IaaS from the providers to meet their end users’ expectations [10]. In fact, data security and privacy are highly important in the context of Cloud 2.0 where cloud, fog and IoT must be consolidated and application providers are granted privileges to use and process the data [11]. In this context, to ensure the availability and delivery of low-latency services, Cloud 2.0 can be integrated with fog and edge computing to deal with the massive data volumes being produced by devices and users [12].

As an example of a cloud provider moving towards this revolution, SITA2 is an IaaS provider that offers mobility-friendly on-demand hosting and application services specifically designed for the air transport industry. SITA has connected more than 160 airports which enabled the organization to host applications accessing to airports systems, such as terminals, gates and parking. A research conducted by Microsoft Cloud and Hosting Study3 also confirmed the Cloud 2.0 movement by showing that 89% of companies are willing to pay additional fees for cloud management services. Despite the large number of pricing competition models, to the best of our knowledge, no one tackled the issue of the cloud providers competition from the perspective of service quality and end-users satisfaction. The only study about quality competition has been conducted by Fan et al. [13] who considered market competition among a software as a service provider and a traditional software provider. Their research focus on marketing advantages of bundling software in a service, neglecting the tight competition among cloud providers themselves and the user satisfaction effect on providers’ revenue.

Considering the initiatives of Cloud 2.0 movement, this paper promotes a healthy market competition through rigor economical and theoretical models. To build a practical roadmap, we propose to empower new and small providers by considering two key features of Cloud 2.0:

  • 1.

    High quality services: Considering the increasing number of clouds deployed in private data centers, the classic approach, such as the one used by Amazon, to build a cloud in which hardware and software developments are insourced, is no longer efficient and hardly deployable. Instead, clouds are being built out of commercial technology stacks with the aim of enabling the infrastructure providers to access the market rapidly and compete while providing high-quality services. However, finding cost-efficient component technologies offering high reliability, continues support, adequate quality, and easy integration is highly challenging. Unlike most of the research performed in the cloud market competition focusing only on pricing mechanisms, we model the competition from the perspectives of cloud service quality and user’s satisfaction by focusing on added-value and superior quality services. Enabling small or new providers to access the market and offer personalized added-value services within our proposed model is part of this feature of Cloud 2.0 that enhances compatibility with private clouds.

  • 2.

    Long-term commitment: The success of modern business applications relies on the reliability of services, such as incident response, security hardening, SLA assurance, software updates, and performance tuning. In fact, 80 percent of downtime is caused by service provisioning problems. Traditionally, these services have been delivered by the IT departments, and simply deploying remote servers in the cloud does not solve the services problem. Because services in the cloud will most likely be outsourced, they must be delivered while considering the customer’s needs in a long-term commitment vision. Moving towards this long-term commitment strategy will drive providers to better focus on customer satisfaction to enjoy higher benefits. Our simulations also confirmed that providing added-value services along with customization could increase long-term commitment which is indeed very profitable.

In this paper, we consider the problem of IaaS cloud market share taking into consideration the need for new cloud providers to be in the market and the requirements of Cloud 2.0. We propose a conceptual, user-centric two-stage game theoretical framework that can help the IaaS providers and users optimize the service quality with a balanced profit. The first stage of our conceptual framework uses our Stackelberg game [14] to identify the user demand preferences and set the optimal price and capacity for the IaaS provider. The Stackelberg game model focuses on interaction among a single IaaS provider with a group of users to appropriately capture the demand elasticities and set the price and allocate resources for each Virtual Machine (VM) to meet the Service Level Agreement (SLA) and match the customers interests. However, our Stackelberg model does not consider the competition among the IaaS providers to provide higher quality services. Therefore, in the second stage, we formulate this competition through a differential game with service quality features as the main competitive factors. Most of the studies on strategic interactions among the cloud participants are grounded in static frameworks [15], [16]. These models overlook the strategic issues arose when providers interact repeatedly over time. Thus, to tackle the limitation of static frameworks, we introduce a non-cooperative dynamic differential game that captures the important dimension of time.

The designed differential game takes multi-tenancy property into account, which leads to define competitive advantages for both the large and small IaaS providers. The large providers (the market leaders) make their profit through a virtuous cycle reflected through the following causal associations: (1) the more customers an IaaS provider gets, the more infrastructure and the better resource provision with robust cloud features (e.g., higher availability and more storage) it can afford; (2) the more infrastructure, the better economies of scale and the cheaper prices for IaaS; and (3) the lower prices and the better their quality, the more customers the provider can get. Meanwhile, the small IaaS providers have fewer users and limited resources. Thus, by targeting a specific industry or local region, they can have tenants who share the same scheme with similar requirements such as complying with data and security regulations, national and international standards or dealing with compatibility issues. This enables them amalgamate their needs by customizing their services to add value to the users’ business solutions. Providing personalized cloud services can further drive customer loyalty [17]. To reflect the above arguments and take them into account, we introduce three main competition factors including ratings by users that reflect customers satisfaction, low cost QoS provisioning, and customization or added-value services.

In summary, our main contribution is a two-stage game theoretical model that:

  • Allows new and small IaaS providers to compete against the existing and large ones and have a market share, which enables a productive cloud market industry that benefits the end-users. To the best of our knowledge, our work is the first that investigates this competition in the cloud computing context.

  • Maximizes users satisfaction modeled using users’ ratings by providing a continues service quality development. It is the first research that models a dynamic competition considering the quality of service among IaaS providers.

  • Captures user preferences and demand elasticities for optimal price and resource allocation. To ensure the continued validity of the optimality in the presence of changing internal or external factors, a post-optimality analysis is provided.

The proposed model can help new born IaaS providers identify their users’ needs and potential markets, anticipate their competitive advantage, formulate their valuation model and create new service provisioning scenarios. We implement our model using a real-world dataset containing users’ ratings over cloud service quality features, obtained from the CloudArmor project.4 Finally, it is worth mentioning that because the problem of making the cloud market competitive by analyzing how small providers can get a market share has not been addressed yet, no benchmark has been found for the purpose of comparison.

Section snippets

Related work

Small and medium businesses can take the advantage of cloud computing in several ways [18]. Cloud computing offers scalable services that businesses can use on demand as much as they need to. The competitive market of cloud services provides a variety of options in pricing and quality. The users can always shift their host provider to another provider offering more opportunistic service or lower price. Due to this opportunistic characteristics, this industry is predicted to reach $270 billion

Framework overview

The race to maximize the revenue, specifically for the new entrants to the cloud market, entails formulation of non-cooperative games. We form two key competing players representing each a group of the same type: (1) a small and fresh provider, and (2) a large and reputed provider. Conventional game theoretic frameworks modeling competition among players highlighted static models. A dynamic model enables the dimension of time, highly important to recognize the competitive decisions that change

Remark

In concrete cases, IaaS providers have multiple types of VMs with different prices and performances. However, this situation does not limit the applicability of our model. We only considered one type of VM for the sake of simplicity and presentation in our mathematical modeling. Since there is no interdependence of VM factors in our model, a real IaaS provider who provides more than one type can use the model for every single type of VM considering the competitors who provide the same type.

IaaS selection Stackelberg game

In the first game, we formulate a Stackelberg game that models the cloud service market interactions between a typical IaaS provider as a leader and the service users as followers. The users observe the price and ratings to adjust their demand accordingly. In quest of the users’ demands, the IaaS provider makes a decision on its pricing strategy and optimal capacity. The game parameters are provided in Table 1.

We define the user demand as a Cobb–Douglas function to model demand elasticities and

Differential competition game

In the previous section, we formulated a static game involving a typical IaaS provider and a typical IaaS user. However, the IaaS provider does not act alone in the market. After identification of the users’ requirements, the IaaS provider needs to plan a suitable strategy against its competitors. To model dynamic competition among the IaaS providers over a period of time, we design a differential game with continuous strategies over a finite horizon time T. The game is between a typical new

Post-optimality analysis

The input data in theoretical optimization approaches is not subject to change, however, in real life it might be found impractical. This assumption is rather valid in a static and deterministic environment, while the essence of our problem is dynamic. User demand reflects market behavior that is changing, and in some degree unpredictable. Cost and capacity estimates are sometimes prone to errors and to changes over time due to the dynamic behavior of the market. Therefore, an important

Experiments and analysis

As the main purpose of our experiments is to demonstrate the effectiveness of the proposed games, we have to set meaningful data and reasonable game parameters. To do so, we obtained real-world data and previously achieved suitable values for the parameters of the Cobb–Douglas demand function [14]. Initially, we experimented with 300 IaaS users for the small provider k1 using real customer ratings to investigate the sensitivity of pricing formula to VM request arrival rate. The data was

Conclusion

This paper tackled the issue of oligopoly IaaS market that neglects the user satisfaction and threatens the grows of the cloud market industry. A conceptual game theoretical framework with two games, namely Stackelberg and differential has been introduced and designed to allow new and even small IaaS providers to obtain a market share using their own strategic advantages. The theoretically obtained results were confirmed by experiments using real-world dataset. It was found that the user demand

Future work and research directions

Although game theory has been applied to real problems in different domains such as political science, biology, and engineering, the assumption that players are rational and have common knowledge so they aim at maximizing profit and minimizing cost is not always practical as shown by some experimental studies [42]. These experimental proposals demonstrated that in some cases, players consider in their decision-making other preferences than simply maximizing profits, for instance psychological,

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.

Acknowledgment

We would like to thank Natural Sciences and Engineering Research Council of Canada (NSERC) for their financial support including Discovery Grant for Jamal Bentahar and Vanier Canada Graduate Scholarship for Mona Taghavi. We would also like to thank the Center on Cyber-Physical Systems at Khalifa University, Abu Dhabi.

Mona Taghavi is a Ph.D. Candidate at Concordia Institute for Information System Engineering, Concordia University, Canada. She obtained her M.Sc. in management information systems from University Kebangsaan Malaysia in 2012 and BSc. in information technology engineering from Islamic Azad University, Iran in 2008. Her research interests include services computing, game theory, blockchain, machine learning and recommender systems.

References (46)

  • DuanW. et al.

    Do online reviews matter? An empirical investigation of panel data

    Decis. Support Syst.

    (2008)
  • ShenB. et al.

    Profit optimization in service-oriented data market: A stackelberg game approach

    Future Gener. Comput. Syst.

    (2019)
  • HabibiM. et al.

    Efficient distribution of requests in federated cloud computing environments utilizing statistical multiplexing

    Future Gener. Comput. Syst.

    (2019)
  • MisraS.C. et al.

    Identification of a company’s suitability for the adoption of cloud computing and modelling its corresponding return on investment

    Math. Comput. Modelling

    (2011)
  • MukhopadhyayS.K. et al.

    A dynamic model for optimal design quality and return policies

    European J. Oper. Res.

    (2007)
  • ZhangF. et al.

    Online learning offloading framework for heterogeneous mobile edge computing system

    J. Parallel Distrib. Comput.

    (2019)
  • ChenJ. et al.

    Tradeoffs between profit and customer satisfaction for service provisioning in the cloud

  • BestavrosA. et al.

    Toward an open cloud marketplace: Vision and first steps

    IEEE Internet Comput.

    (2014)
  • DurkeeD.

    Why cloud computing will never be free

    Commun. ACM

    (2010)
  • XuB. et al.

    Optimal pricing for the competitive and evolutionary cloud market.

  • WillcocksL.P. et al.

    Moving to the Cloud Corporation - How to face the challenges and harness the potential of cloud computing

    (2014)
  • HobergP. et al.

    The business perspective on cloud computing - a literature review of research on cloud computing

  • SuleimanB. et al.

    On understanding the economics and elasticity challenges of deploying business applications on public cloud infrastructure

    J. Internet Serv. Appl.

    (2012)
  • Cited by (12)

    • Cooperative reliability allocation in network flow problems considering greenhouse gas emissions: Optical fiber networks structure

      2021, Journal of Cleaner Production
      Citation Excerpt :

      Competition between Infrastructure as a Service (IaaS) providers are also considered with reference to cloud market share (Taghavi et al., 2020); market dominance by certain providers may obstruct healthy competition, thus causing extra costs to users. Taghavi et al. (2020) argued that provisions should be made to enable new providers to enter the market in order to increase competition. So far, though, the effect of CGT on increasing reliability when designing the ICT sectors of network utilities has not been considered.

    • Resource management in the federated cloud environment using Cournot and Bertrand competitions

      2020, Future Generation Computer Systems
      Citation Excerpt :

      Many studies have been done in this field, and they have introduced different models and algorithms for this problem [5–8]. Game-theory introduces a framework for situations of conflicting interests and has attracted lots of researchers in the area of cloud computing in recent years [9,10]. Game-theory is very suitable for the situation that we have the number of independent players that their decision affected by interactions of others’ behavior with their own.

    • Optimal pricing of subscription services in the restaurant industry

      2024, Journal of Revenue and Pricing Management
    View all citing articles on Scopus

    Mona Taghavi is a Ph.D. Candidate at Concordia Institute for Information System Engineering, Concordia University, Canada. She obtained her M.Sc. in management information systems from University Kebangsaan Malaysia in 2012 and BSc. in information technology engineering from Islamic Azad University, Iran in 2008. Her research interests include services computing, game theory, blockchain, machine learning and recommender systems.

    Jamal Bentahar received the bachelor’s degree in software engineering from the National Institute of Statistics and Applied Economics, Morocco, in 1998, the M.Sc. degree in the same discipline from Mohamed V University, Morocco, in 2001, and the Ph.D. degree in computer science and software engineering from Laval University, Canada, in 2005. He is a Professor with Concordia Institute for Information Systems Engineering, Faculty of Engineering and Computer Science, Concordia University, Canada. From 2005 to 2006, he was a Postdoctoral Fellow with Laval University, and then NSERC Postdoctoral Fellow at Simon Fraser University, Canada. He is an NSERC Co-Chair for Discovery Grant for Computer Science (2016–2018). His research interests include the areas of cloud and services computing, game theory, computational logics, model checking, multi-agent systems, and software engineering.

    Hadi Otrok received his Ph.D. in ECE from Concordia University. He is an Associate Professor in the Department of ECE at Khalifa University. He is an associate editor of the Ad-hoc networks, IEEE communications letters and a co-chair of several committees at various IEEE conferences. His research interests include services computing, network security, game theory and mechanism design.

    View full text