Cost model based service placement in federated hybrid clouds

https://doi.org/10.1016/j.future.2014.08.014Get rights and content

Highlights

  • Comparison of cost models for cloud computing.

  • Design of a comprehensive cost model for federated hybrid clouds.

  • Design of an algorithm for service placements in federated hybrid clouds.

  • Performed a sensitivity analysis with the help of the service placement algorithm.

Abstract

As cloud federation allows companies in need of computational resources to use computational resources hosted by different cloud providers, it reduces the cost of IT infrastructure by lowering capital and operational expenses. This is the result of economies of scale and the possibility for organizations to purchase just as much computing and storage resources as needed whenever needed. However, a clear specification of cost savings requires a detailed specification of the costs incurred. Although there are some efforts to define cost models for clouds, the need for a comprehensive cost model, which covers all cost factors and types of clouds, is undeniable. In this paper, we cover this gap by suggesting a cost model for the most general form of a cloud, namely federated hybrid clouds. This type of cloud is composed of a private cloud and a number of interoperable public clouds. The proposed cost model is applied within a cost minimization algorithm for making service placement decisions in clouds. We demonstrate the workings of our cost model and service placement algorithm within a specific cloud scenario. Our results show that the service placement algorithm with the cost model minimizes the spending for computational services.

Introduction

The cloud computing paradigm has been established as a promising model for using computational resources and software resources on-demand  [1], [2], [3]. As one of its advantages, it shifts complex IT resource management tasks typically performed by a company that is in need of computational resources to service providers. It also increases the flexibility of companies to adapt their IT resources to changes in demand  [4], [5]. Assuming a cost-based pricing of cloud services, it can reduce infrastructure investment cost and cost for organizing IT resources. The capital expenditure (CAPEX) and operational expenditure (OPEX) are lower due to the economies of scale of the cloud providers’ infrastructure.

A potential scenario for the foreseeable future is that of a hybrid cloud environment. Hybrid clouds refer to a composition of a private cloud (i.e., data center managed through cloud technology and owned by the company in need of computational resources) and at least one public cloud (i.e., resources owned by a cloud provider)  [6]. Most enterprises, except for those with a very small or stable demand for computing resources, will employ a mix of in-house and outsourced computing resources  [7]. The reason for keeping a private cloud, despite the economies of scale of public clouds, is the increased security risks in the public cloud and possibility of serving the base demand at lower cost  [8]. Therefore, with the ability of a seamless migration of processing load between a private cloud and public clouds, the hybrid cloud is ideal for handling variable demand  [9].

To truly fulfill the promise of cloud computing (i.e., its benefit from economies of scale), technology is needed to federate disparate public clouds at the infrastructure level (IaaS). Only through federation and interoperability at this level, small cloud providers (i.e., providers of limited scale) can take advantage of their aggregated capabilities and provide a seemingly infinite service computing utility that can compete with the bigger cloud providers in the market. We refer to the infrastructure that supports this paradigm as federated clouds  [10]. By definition, a federated cloud is a cloud, in which (competing) cloud providers have reached a cross-site agreement for cooperating regarding the deployment of service components in a way similar to electrical power providers who use capacity from each other to cope with demand variations among their own customers  [11].

To take advantages of hybrid federated clouds, companies need to know the overall cost of running their services on this kind of cloud infrastructure  [12], [13], [14], [15]. A service runs on one or more virtual machines (VMs) and can be moved (with the VMs) between federated clouds. The reports on the cost savings of clouds vary, however. For example, a report by McKinsey states that moving to the cloud would actually cost 144% more than the current data center infrastructure  [9], [16]. West reports that cloud technology reduced a government agency’s cost by 25%–50%  [16]. This uncertainty about cost savings and about actual expenditure were also discussed by Kondo et al.  [4]. These examples highlight the necessity for an overall cost model.

Apart from the cost model for federated hybrid clouds, the service placements on this kind of cloud infrastructure need to be considered by the company that is in need of computational resources, as each service placement incurs a different cost. Hence, minimizing the overall cost for using federated hybrid clouds requires an optimization model that considers all cost factors and cost values for each of the service placements  [17], [18]. Whenever the configuration and prices of clouds change over time, the optimization problem has to be solved again  [11], [14], [19].

Summarizing this discussion, our research can be described with two research questions: (1) What are the cost factors specific to federated hybrid clouds and how can a comprehensive cost model for federated hybrid clouds be defined? (2) If such a cost model for federated hybrid clouds exists, how can it be used for a service placement optimization algorithm?

In particular, we intend to construct an overall cost model that can be used by enterprises to decide where to place their services on a federated hybrid cloud such that it minimizes cost. This overall cost model, which is based on the initial and less comprehensive version of the cost model given in  [20], comprises all cost factors and cost functions, which are necessary to estimate the precise cost of running cloud services on an in-house cloud-enabled data center (private cloud) and on federated public clouds.

To answer the research questions, we conduct the following steps: First, we perform a systematic literature review of papers on cost factors and cost models in cloud computing. Second, we identify the gaps in the current research on federated hybrid cloud cost modeling. Based on this result, we design a comprehensive overall cost model for federated hybrid clouds. In the next step, we propose an optimization algorithm for service placement on federated hybrid clouds, using this cost model. Finally, we apply the proposed cost model and the service placement optimization algorithm to a case study to demonstrate its workings.

The contributions of this paper are threefold: First, it comprises a comprehensive cost model that cannot only be used for cost calculation of federated hybrid clouds (as in this paper) but also to guide and compare investment decisions in private clouds. Second, the cost model considers cost for the migration between clouds which includes the deployment cost and the transmission of data between public clouds (and not just between private clouds and public clouds). Third, a brute-force service placement algorithm that considers the cost of service placements.

The remainder of the paper is organized as follows. The next section gives an overview about cost factors and cost models for cloud computing, which have been proposed in literature. Section  3 introduces our cost model for federated hybrid clouds. In Section  4, our cost-model-based service placement optimization algorithm is presented. Furthermore, an application of our cost model as part of the service placement algorithm to a cloud computing scenario is shown. Section  5 concludes the paper with a brief discussion.

Section snippets

Cloud computing, hybrid clouds, and federated clouds

Although there are many definitions of cloud computing, the NIST definition seems to have captured the commonly agreed cloud computing aspects that are mentioned in most of the academic papers  [5]. The NIST definition states that cloud computing is “A model for enabling convenient, on-demand network access to a shared pool of configurable computing resources (e.g., networks, servers, storage, applications, and services) that can be rapidly provisioned and released with minimal management

User application model

Our user application model assumes a service-oriented architecture. A user application, which represents a business process, comprises the execution of M services. Services are assumed to be basic units, which can communicate with each other by transferring a specific amount of input data and output data. Each service runs on a VM. An example of this user application model is shown in Fig. 2.

The representation of services and the data transfer relationships between them through graph theory

Service placement optimization algorithm

Finding the optimal service placement requires the calculation of the total cost for each possible service placement option on a private cloud and a number of federated public clouds. To achieve this, our cost model based service placement optimization algorithm, which is called COMBSPO, initializes the main input variables (Section  2.4) and, then, executes a loop over all possible service placements. Within the loop, the calculation of the total cost (Eq. (6)) is performed based on our cost

Conclusion and future work

In this paper, we have comprehensively analyzed the cost factors that need to be considered for estimating cloud computing cost of federated hybrid clouds, i.e., the most general cloud architecture. As part of this work, the cost factors have been evaluated and categorized. This analysis has also been the basis for our cloud cost model, which can be used by any cloud customer to estimate the total cost of clouds (i.e., private clouds, public clouds, hybrid clouds, distributed clouds, federated

Acknowledgment

This work has been funded by the Korea Institute for Advancement of Technology (grant no. ITEA2 10014) as part of the ITEA2 project EASI-CLOUDS.

Jörn Altmann is Professor for Technology Management, Economics, and Policy at the College of Engineering, Seoul National University. Prior to this, he taught computer networks at UC Berkeley, worked as a Senior Scientist at Hewlett-Packard Labs, and has been a postdoc at EECS and ICSI of UC Berkeley. Dr. Altmann’s research centers on Internet economics with a focus on economic analysis of Internet services and on integrating economic models into Internet infrastructures.

References (51)

  • R. Van den Bossche et al.

    Online cost-efficient scheduling of deadline-constrained workloads on hybrid clouds

    Future Gener. Comput. Syst.

    (2013)
  • B.C. Tak, B. Urgaonkar, A. Sivasubramaniam, To move or not to move: The economics of cloud computing. HOTCLOUD 2011,...
  • M. Klems, J. Nimis, S. Tai, Do clouds compute? A framework for estimating the value of cloud computing. Designing...
  • A. Bany Mohammed et al.

    Cloud computing value chains: understanding businesses and value creation in the cloud

    Econ. Models Algorithms Distrib. Syst.

    (2010)
  • D. Kondo, B. Javadi, P. Malecot, F. Cappello, D.P. Anderson, Cost-benefit analysis of cloud computing versus desktop...
  • P. Mell et al.

    The NIST definition of cloud computing

    Natl. Inst. Stand. Technol.

    (2009)
  • C.D. Patel, A.J. Shah, Cost Model for Planning, Development and Operation of a Data Center. HPL-2005-107(R.1). June...
  • A. Cole, The future belongs to hybrid clouds. IT Business Edge. September 2011,...
  • G. Orenstein, Forecast for 2010: The rise of hybrid clouds. GIGAOM. January 2010....
  • A. Cole, Hybrid Clouds Allow Organizations to ‘Own the Base and Rent the Peak’. IT Business Edge. July 2011....
  • B. Rochwerger et al.

    The reservoir model and architecture for open federated cloud computing

    IBM J. Res. Dev.

    (2009)
  • E. Elmroth et al.

    Interfaces for placement, migration, and monitoring of virtual machines in federated clouds

  • E. Deelman, G. Singh, M. Livny, B. Berriman, J. Good, The cost of doing science on the cloud: The montage example. SC...
  • D. D’Agostino et al.

    A QoS-aware broker for hybrid clouds

    J. Comput.

    (2013)
  • S. Pandey, A. Barker, K.K. Gupta, R. Buyya, Minimizing execution costs when using globally distributed cloud services....
  • L. Bittencourt et al.

    HCOC: A cost optimization algorithm for workflow scheduling in hybrid clouds

    J. Internet Serv. Appl.

    (2011)
  • D.M. West

    Saving money through cloud computing

    Governance Studies, Brookings

    (2010)
  • M. Bjorkqvist, L.Y. Chen, W. Binder, Cost-driven service provisioning in hybrid clouds. SOCA 2012, 5th International...
  • Y. Kessaci et al.

    A Pareto-based metaheuristic for scheduling HPC applications on a geographically distributed cloud federation

    J. Cluster Comput.

    (2013)
  • M. Kashef et al.

    A cost model for hybrid clouds

  • J. Metzler et al.

    Cloud Computing: Reality vs. Fiction

    Network World

    (2008)
  • M. Maurer et al.

    Cost and benefit of the SLA mapping approach for defining standardized goods in cloud computing markets

    International Conference on Utility and Cloud Computing

    (2010)
  • J. Altmann et al.

    A marketplace and its market mechanism for trading commoditized computing resources

    Ann. Telecommun.

    (2010)
  • J.F. Rayport, A. Heyward, Envisioning the cloud: the next computing paradigm and its implication for technology policy....
  • J. Weinman, Mathematical proof of the inevitability of cloud computing, White Paper, January 2011,...
  • Cited by (92)

    • Economic model for evaluating the value creation through information sharing within the cybersecurity information sharing ecosystem

      2021, Future Generation Computer Systems
      Citation Excerpt :

      Two-sided network effects explain the double impact of both types of network effects. In traditional markets of software products, for example, the major part of value is conceived through the two-sided network effects [12,173,174]. Specifically, network effects have impact on market competition, equilibrium, compatibility decisions, and on adoption pace of new technologies [173,175–178].

    • A survey study on virtual machine migration and server consolidation techniques in DVFS-enabled cloud datacenter: Taxonomy and challenges

      2020, Journal of King Saud University - Computer and Information Sciences
      Citation Excerpt :

      For instance, Amazon EC2 provisions virtual servers, bandwidth for data transferring and storage in the form of infrastructure as a service (IaaS) (Amazon EC2, 2018), Microsoft Windows Azure provisions Integrated Development Environment (IDE) tools in the form of platform as a service (PaaS) for developers to develop their cloud-based software and publish their productions in the cloud environment (Microsoft Windows Azure, 2018) and Salesforce provides customer relationship management (CRM) software for different industries in the form of software as a service (SaaS) (Salesforce, 2018). The concept of TCO in cloud DCs contains two big categories: CAPEX and OPEX in which the former indicates onetime expenditures such as providing commodity cluster servers, network devices, etc. whereas the latter indicates dynamic expenditures such as electricity costs for electronic and non-electronic devices, human force salaries, real estate/mortgage and etc. (Altmann and Kashef, 2014). Cloud providers try to lower down their OPEX to reach maximum benefit instead of increasing their service tariff.

    • Predictive service placement in cloud using deep learning and frequent subgraph mining

      2023, Journal of Ambient Intelligence and Humanized Computing
    View all citing articles on Scopus

    Jörn Altmann is Professor for Technology Management, Economics, and Policy at the College of Engineering, Seoul National University. Prior to this, he taught computer networks at UC Berkeley, worked as a Senior Scientist at Hewlett-Packard Labs, and has been a postdoc at EECS and ICSI of UC Berkeley. Dr. Altmann’s research centers on Internet economics with a focus on economic analysis of Internet services and on integrating economic models into Internet infrastructures.

    Mohammad Mahdi Kashef is a Ph.D. student at the Technology Management, Economics, and Policy Program of the College of Engineering, Seoul National University. His research interest is service placement in clouds.

    View full text