A framework for studying cloud value systems

Abstract

Cloud computing is about to change the way of delivering IT services to end users. Service providers are required to re-think both their value proposition and value creation processes. We present a framework for studying these cloud value systems. The framework is grounded on New Institutional Economics and provides means for describing, explaining, and designing cloud value systems.

Full text

1
A Framework for Studying Cloud Value Systems
Joerg Leukel, Stefan Kirn
University of Hohenheim, Schwerzstr. 35, 70599 Stuttgart, Germany
joerg.leukel@uni-hohenheim.de, stefan.kirn@uni-hohenheim.de
Abstract
Cloud computing is about to change the way of delivering IT services to end users. Service providers
are required to re-think both their value proposition and value creation processes. We present a
framework for studying these cloud value systems. The framework is grounded on New Institutional
Economics and provides means for describing, explaining, and designing cloud value systems.
Keywords: cloud computing, value system, outsourcing, new institutional economics, coordination
This is the authors’ version of the following journal article:
Leukel, J., & Kirn, S. (2011). A framework for studying cloud value systems.
it – information technology, 53(4), 195-201. http://dx.doi.org/10.1524/itit.2011.0643
1. Introduction
Recent studies indicate that cloud computing will affect fundamentally the way how software
companies provide IT services to end users [6]. For instance, ERP vendors such as SAP have
implemented massive reorganizations to address demands for cloud-based solutions [17]. In this
sense, cloud computing is not limited to the interface of provider and end user, but concerns the
entire value creation process of software companies, including both internal added-value by
production and external added-value by sourcing [12]. The practical implications as well as
potentials of cloud computing on this process have not yet been sufficiently understood. Our research
aims at contributing to this problem by an analytical framework.
We study cloud value systems from the perspective of New Institutional Economics. This set of
theories is adequate for explaining the rules and norms that constrain economic activities. It pays
special attention to the role of information and thus potential changes by adopting IT. The framework
could serve as a general tool for further explaining and ultimately designing cloud value systems.

2
The remainder is structured as follows. Section 2 introduces our conceptualization of cloud
value systems. Section 3 presents the framework and its constituents. Section 4 summarizes the
preliminary findings and points to future work.
2. Cloud Value System
2.1 Definition
In the following, we ground cloud computing on the definition proposed by Vaquero et al. [18],
who analyzed and distilled the essence of twenty-two definitions: A cloud is a pool of computing
resources such as hardware, platforms, and software services, which can easily be used and exploited
by a pay-per-use model. The pool is managed for optimal resource utilization, while guarantees for
usage exist in terms of customized service level agreements (SLA).
With regard to its organization, a cloud typically distinguishes at least two roles, infrastructure
provider (IP) and service provider (SP). While the former provides the actual resources, the latter
delivers end user services by using such resources. Ideally, the two roles are fulfilled by different
organizations; thus the service provider can concentrate its efforts on services without catering for all
implementation issues. Therefore, only the SP interacts with end users.
In a cloud, all resources are provided 'as a service'. Depending on the type of service, three
scenarios can be formed: 'Infrastructure as a Service' (IaaS) is about computing resources, 'Platform
as a Service' (PaaS) is about software platforms, and 'Software as a Service' provides software
applications.
Following this organization, we define cloud value system as shown in definition 1 and figure 1.
Definition 1 (Cloud Value System): Cloud value system CVS=(s, I, U, F, T) consists of a
service provider s, its upstream infrastructure providers I={i
1
, .., i
m
}, and its downstream end users
U={u
1
, .., u
n
}, being connected by service flows f
∈
F, with F
⊆
(I
×
s) UNION (s
×
U). T is a function
T

{IaaS, PaaS, SaaS}, which maps each f to one service type (infrastructure, platform, or software).
The cloud value system is centered on the service provider. This conceptualization is consistent
with value system analysis in economics, because it aims at assisting decision makers of the focal
firm. It has to be noted, though, that the infrastructure providers in our model may maintain their own
cloud value system, in which our I takes over the role of s, our s is an end user, and additional IPs
exist further upstream. This is just the idea of distinguishing roles (instead of actors).

3
s
i
1
i
m
service
flow
.
.
.
.
.
.
u
1
u
n
.
.
.
.
.
.
service
flow
service
flow
service
flow
Figure 1. Cloud value system.
Our conceptualization of IP and SP is, however, both an abstraction and detailling of actual
clouds. First, it abstracts from many more roles that can be identified for clouds, such as by
distinguishing the type of cloud service provided (e.g., platform provider, storage provider), or
separating component services from composite services (e.g., service aggregator). The literature
review in [3] identified 108 different role names and reduced them to a core set of eight roles,
including the customer. Our model can be mapped to five of these roles, when considering the cloud
service type and whether the respective actor provides the service directly to the end user/customer.
The most important difference is that we use the latter as a discriminator for IP vs. SP, whereas
Boehm et al. use the role IP to denote any provider of an IaaS cloud service, regardless of the its
position in the value system.
Second, the distinction of IP and SP does not materialize fully in the cloud market. The big
player such as Amazon, google, IBM, and Fujitsu all can be regarded as full service providers that
integrate both roles in one firm. At least, this is how the customer perceives the cloud service
delivery and thus would regard it as SP only. The reason is, however, that these firms hide the
underlying complexity of cloud service delivery - in our terms: the cloud value system; though it
does exist in any case, either internally (IP roles would be mapped to organizational entities) or as an
inter-organizational system (IP roles would be mapped to other firms).
2.2 Cloud Economics
Cloud computing has attracted the interest of researchers from economics and the IS discipline.
The broad term 'cloud economics' refers to the analysis of production, distribution, and consumption
of services in cloud value systems. Its antecendts can be found in research on prior, related
computing technology such as grid computing, as well as IT outsourcing [4]. The most relevant work
to ours is that by Boehm et al. [3] and Leimeister et al. [10]. They propose a 'cloud computing value

4
network' model. It aims at allowing a systematic description of actors in the cloud computing market,
their interactions, and by thus at improving the understanding of opportunities for such actors.
Whereas this research can be regarded as a comprehensive model for the cloud market and cloud
economics, our research is focused at analyzing the interactions that occur in the three-tier value
system of IPs, SP, and end users. Mohammed et al. [13] also study cloud value systems, but their
proposal entitled 'Cloud Value Chain Reference Model' is actually confined to describing value
creation of one firm, not that of the entire value system. The reason is that the underlying theory is
Porter's value chain.
Next, we describe key changes associated with cloud computing and decisions in firms that are
potentially affected. This outline sets the direction of our framework in section 3.
First, cloud computing promotes the pay-per-use model for computing resources. These
resources are charged similar to utilities based on actual consumption [16] rather than availability or
ownership. There are no capital expenses for service consumers. The effect on costs is a reduction of
fixed costs; in an ideal scenario, fixed costs are zero and all other costs are variable [7]. This change
takes place for single firms that use cloud services as well as for entire industries, because of
increased resource sharing, pooling, and efficient work balancing [12]. The technical solution is
virtualization, thus the decoupling of applications from (physical) infrastructure. The shift from fixed
to variable costs, however, is only one side of the same coin: Any provider is faced with an
increasing share of variable revenue replacing fixed revenue.
Any cost change of an input factor (i.e., resource used) has to be considered in a firm's decision
making. The expected shift from high fixed costs and long-term investments to variable costs and
short-term contracts necessarily requires to re-thinking make-or-buy decisions in firms. These
decisions weight the benefits and risks of either producing a good/service by the firm or buying it
from a third party. Concerning the cloud value system, IPs need to assess the possibility to deliver
services to its end users by sourcing from IPs.
Second, cloud computing is attributed with better balancing demand for and supply of resources
(i.e., by virtualization and efficient resource sharing across applications and industries) [18]. For
instance, resources can be dynamically re-configured to adjust to a variable load. In a cloud value
system, the SP maintains two respective views: As a provider of services to end users, he aims at a
maximum of service delivery which makes up his revenue stream. As a buyer of services from I, he
aims at minimal costs while meeting the SLAs of his customers.

5
Since cloud computing promotes a highly dynamic supply-side, reconciling demand and supply
is a critical task for any SP [9]. The question is how to manage the dependency between incoming
and outgoing service flows. The decision problem is that of designing an efficient governance.
3. Framework
3.1 Overview
We propose a framework for studying cloud value systems. The overall objective is to provide a
theoretical base for describing, explaining, and designing cloud value systems. More specifically, it
should help decision makers in decisions about (1) make-or-buy and (2) governance of 'buy'-
relationships. We employ New Institutional Economics as the overarching body of theories. These
theories study the rules and norms that constrain economic activities. They have enjoyed wide
acceptance in both economics and IS research to explain potential changes that may occur by
adopting new information technology. In particular, these theories are helpful for analyzing IT
outsourcing decisions [1; 14].
Figure 2 gives an overview of the analytical framework consisting of three general theories:
transaction cost, agency, and property rights theory.
Cloud Value System
(CVS)
Transaction Cost Theory
Perspective:
CVS as a system of economic
transactions of service flows
Problem:
Optimal coordination mechanism
(minimal costs)
Agency Theory
Perspective:
CVS as a system of principal-
agent-relationships
Problem:
Optimal contract between
principal and agent
Property Rights Theory
Perspective:
CVS as a system of contracts defining
property rights of service flows
Problem:
Optimal distribution of property rights
between I, s, and U
Figure 2. Perspectives and problems of Cloud value systems in New Institutional Economics.

6
3.2 Transaction Cost Theory
Transaction cost theory (TCT) studies the costs incurred in making an economic exchange [19].
A transaction transfers a good/service as well as so called property rights on the good/service. For
this purpose, a transaction is divided into phases, each incurring specific costs. These phases are
information (i.e., searching for and comparing offerings), settlement (i.e., bargaining an agreement
between seller and buyer), and enforcement (i.e., monitoring the delivery and assuring its agreed
quality). Transaction costs are thus made of information costs, settlement costs, and enforcement
costs. They are separated from production costs. The goal of individual actors is to minimize total
costs. TCT contributes a set of determinants of transaction costs (specificity, frequency, and
uncertainty). The theory can be used to decide about the optimal coordination mechanism, thus the
one which incurres minimal transaction costs. Next, we adopt the findings of TCT to cloud value
systems.
Scope: The concept of transaction matches with service flow. Each service flow incurs costs at
the buyer. In our SP-centered analysis, transactions between SP and IP are most relevant, because the
SP needs to decide about the design of such transactions.
Transaction phases: Searching for cloud services and providers of either type (IaaS, PaaS,
SaaS) causes search and information costs. This task may include manual search (personal costs of
staff) or automatic search by means of service discovery and selection, the latter incurring marginal
variable costs only. Negotiating a service contract ranges also from inter-personal negotiation to
electronic settlement, with respective differences in incurred costs. Enforcement costs are related to
SLA monitoring, service accounting, and service billing. Despite significant advances in the areas of
service description, discovery, negotation, and SLA management, cloud computing still suffers from
a lack of standardization and consolidition of methods and respective software frameworks; the
reason is that service-oriented computing (SOC) has not yet reached the level of maturiy as expected
by SOC research [15].
Specificity: If an asset used as input by a firm is specific to a single purpose, its usage for an
alternative, second best purpose incurs additional costs. Any asset of high specificity therefore
represents a risk for the firm. Assume that the SP buys a service from an IP to meet an individual
request from an end-user. If the end-user cancels the contract or does not consume the requested
service, the SP is in danger of a loss. In general, three types of specificity exist:
(1) Site specificity incurs costs for moving the asset from one physical location to another. Since
cloud services are provided electronically, a second best usage does not require to move the firm;
hence no relevance of site specificity.

7
(2) Physical asset specificity incurs costs for deploying a machine or tool in another
environment for another purpose. This specificity is relevant to cloud: Assume that the SP buys a
PaaS-type service from an IP and deploys a custom-made SaaS-type service. If the respective end
user drops out, then the platform service could be used for another software service, if it can be
deployed on that platform. In general, IaaS-type services are less physically specific than PaaS, and
the latter less than SaaS. This order of physical asset is grounded on important insights of software
engineering and software reuse research [8]; the layer structure of IaaS, PaaS, and SaaS can be
mapped to common levels of software architectures that foster reusability, i.e., operation
system/IaaS, domain-independent software/PaaS, and applications/SaaS.
(3) Human capital specificity is about human skills that can not be used for another purpose. Let
us assume that the SP hires qualified staff for delivering a Windows Server 2008 Datacenter service,
because he expects to acquire new customers in this field. If the customer base does not materialize,
the staff may be used for the general Windows Server 2008 service, but at the higher cost/salary of
Datacenter only. Therefore, human capital specificity has to be considered in the value proposition of
the SP. The order of human capital specificity can also be traced back to layers of software
architecture and software reuse (see above).
Time specificity: A fourth type of specificity was proposed by Malone et al. [11]. The rationale
is that the value of a service is dependent on reaching the user within a certain period of time. We
adopt it to cloud services, which need to be delivered at a specific time respectively over a period of
time. Any failure incurs costs on the provider. Current cloud service offerings reflect the relevance of
this type of specificity in terms of SLAs that include service availability guarantees.
Frequency: If a transaction occurs between two participants often, then economies of scale and
organizational learning can result in decreasing transaction costs over time. As a consequence, firms
are more eager to integrate the transaction vertically, thus switch from market to hierarchy, or to in-
source the service by a long-term contract which avoids information and settlement costs per
transaction. Assume a cloud service for which the SP receives very little demand from end-users; the
SP will not deliver this service. At the most, he will forward the demand to one of his IPs. Assessing
the frequence of cloud services is hardly possible and can not directly backed up by a theory.
Empirical data on the number of transactions over is at most available for few of the horizontal cloud
services, though does not allow to construct a general characterization, in particular ordering the
service types.
Uncertainty: It relates to the number and degree of unforeseen changes during service delivery.
In general, hierarchies are in favor of markets, because they allow a faster identification and reaction

8
to changes. However, bounded rationality limits the ability to react in an appropriate way (e.g.,
required expertise and experiences with the technology used). Again, IaaS is associated with lower
uncertainty than PaaS and SaaS, because of lower complexity of the service system (due to the fact
that PaaS necessarily uses an infrastructure/machine, and SaaS uses a platform/operating system
etc.). Both service description and service delivery of IaaS-type services grounds on a higher level of
standardization, thus less diversity, with contributes to reducing unforeseen changes and thus risks.
The brief analysis of TCT determinants on transactions costs in cloud value systems has shown
that the three types of service flows differ largely. The findings are summarized in table 1.
Table 1. Service flows by TCT determinants.
Type of service flow from IP to SP
Determinat IaaS PaaS SaaS
Site specificity None None None
Physical asset specificity Low Medium High
Human capital specificity Low Medium High
Uncertainty Low Medium High
3.3 Agency Theory
Agency theory is concerned with problems that occur when a principal hires an agent under
conditions of incomplete and asymmetric information as well as opportunistic behavior [5]. In a
CVS, a SP represents the principal in a principal-agent relationship with an IP. Subject of this
relationship is the delivery of a cloud service of either type.
Basically, two problems exist. First, principal and agent aim at conflicting goals; while the
principal can not fully verify that the agent behaves as agreed upon ('agency problem'). Second,
principal and agent prefer different actions to reduce respectively prevent a shared risk ('risk
problem').
The agency problem in our case is that of conflicting goals towards resource usage: The
principal does not cater for efficient resource usage at the agent's site as long as its service requests
are fulfilled, whereas the IP's value system is dependent on resource sharing. As a result, the
principal is in danger of an overselling agent, who speculates for a lower than agreed upon resource
demand over time. This agent is risk-seeking. If the agent is risk-neutral, he would balance its service
offering with available resources; if the agent is risk-averse, he would pursue under-selling of

9
resources. The agency problem causes agency costs. For instance, the principal can improve the level
of control of the agent only by additional measures such as monitoring, reporting, quality inspection
etc. A recommendation for IPs would be to assess thoroughly the agent's characteristics and behavior
prior to agreeing upon a service contract; the agent, however, will unlikely disclose this behavior in a
technical service description or SLA.
The risk problem is that of different strategies to handle a shared service delivery risk. Risk
sharing means that a risk does not affect either the principal or agent, but both. The concrete form of
sharing is necessarily specified in the service contract, e.g, by a payment scheme. Let us assume the
risk of network node failure at the IP and an agent, who in case of risk materialization could
outsource the service to another network node. The agent, however, will follow this strategy only, if
the principal pays an additional fee. Another type of risk is that of volatile end user demand, so that
the IP will request an uncertain service load; this risk could be covered in a mid- or long-term
contract. Again, an agent aims at a guaranteed minimum payment per period to prevent a loss due to
reserved, though unused resources.
3.4 Property Rights Theory
Property rights define or delimit the privileges granted to individuals of specific resources (note
that resource is an economical term here). In this sense, any economic system can be regarded as a
system of such rights; a transaction transfers property rights between two economic parties. Here, we
are concerned with property rights owned and transferred by SP and IP. The question is which rights
exist and how they should be distributed among participants of a CVS.
The theory distinguishes four categories of property rights. These need to be adjusted to CVS in
which property rights are assigned to cloud services of either type. The difference to resource is that
the right is not associated with a specific instance (e.g., machine), but capability. This
conceptualization affects the definitions: Usus is the right to use the cloud service. Usus fructus is the
right to appropriate the returns of the cloud service. Abusus is the right to change the cloud service.
Ius abutendi is the right to sell the cloud service.
With regard to the SP, an Usus property right is not sufficient, but requires at least Usus fructus
(e.g., exploiting a bought PaaS service commercially). This restriction points to combinations of
property rights. The value of a cloud service is made up of its property rights. However, the optimal
allocation of property rights is minimal in this sense, that removing a right would reduce the value.

10
Depending on the service type, property rights concretize only partially due to the nature of
services vs. goods (Table 2).
Table 2. Categorization of property rights in Cloud value systems.
Property right IaaS PaaS SaaS
Usus Usage of cloud services by IP and end users.
Usus fructus Machine/platform used by the for
1. reselling this service
2. or deploying offered software services.
Software service for
resale or integration in
other offered software
services.
Abusus Cloud services cannot be changed in ‘substance’.
Ius Cloud services cannot be sold ‘as a material’.
A property right is not necessarily exclusive (e.g., more than one end user of the same service;
an IP offers a service to multiple SPs). The ultimate goal is to allocate all property rights efficiently,
so that the total utility of all participants is maximized.
4. Conclusion
This paper proposed an analytical framework for studying cloud value systems. By grounding it
on a set of theories of New Institutional Economics, it allows describing economic transactions,
principal-agent relationships, and property rights in the specific setting of cloud computing. The
expectation is that this theoretical foundation could ease the effort for describing, explaining, and
finally designing CVS. As a preliminary finding, we provided some insights into cloud transactions
and pointed to potential cloud governance structures.
The main limitation of this framework concerns its validity. We have not yet applied the
framework to existing cloud value systems, but tried to derive its structure from some theories
(deduction) and illustrate the applicability by some examples. A solid empirical foundation is
missing. Another important issue is related to the service types of IaaS, PaaS, and SaaS. While they
reflect some consensual classifications, they are too abstract for the purpose of applying, e.g.,
transaction cost theory. Cloud services of the same type can differ in important properties such as
specificity (e.g., domain-independent SaaS such as messaging vs. domain-specific SaaS such as
production planning). Being a framework, future research is required in terms of formal constructs
and models, deduction of hypothesis, and empirical studies to both validate the constituents and

11
derive concrete design recommendations. With regard to the transaction theory, the study by Benlian
[2] suggests that specificity of SaaS-typed services is a strong factor for outsourcing. Similar
empirical evidence for other cloud services, though, does not exist yet.
With regard to the agency theory, the focal service provider of a cloud value system has actually
two roles: First, he is the agent in the relationship with his end users. Second, he is the principal in
the relationship with his IPs. Studying the SP thus requires to acknowledge the existence of two
agency relationship, studying both separately, as well as the interdependencies of both. This subject
has yet not been studied in cloud economics, but could learn from similar interdepencies in multi-tier
supply chains, for instance from supply chain management literature. Since agency theory aims at
optimal contracts between principal and agent, a study has to assess the structure and content of
SLAs and its definition of certain Quality-of-Service (QoS) levels. In particular, it could help
deciding about payment schemes respectively fees. For instance, it should be possible to making
shared risks explicit.
With regard to the property right theory, the technical means for describing and transferring
property right are SLAs and QoS parameters. Additional work is required to assess the relationship
between this theory and the technical means provided by SOC, in particular specifications of SLAs
and methods used in SLA negotiation and enforcement.
Acknowledgement: The work presented in this paper was partly funded by the German Federal
Ministry of Education and Research under the project InterLogGrid (BMBF 01IG09010E). We wish
to thank Marcus Mueller and Daniel Weiss for their helpful comments on an earlier version of this
paper.
References
[1] B. Bahli and S. Rivard. The information technology outsourcing risk: A transaction cost and
agency theory-based perspective. In: Journal of Information Technology 18:211-221, 2003
[2] A. Benlian. A transaction cost theoretical analysis of Software-as-a-Service (SaaS)-based
sourcing in SMBs and Enterprises. in: Proceedings of the 17th European Conference on
Information Systems (ECIS 2010), Verona, Italy, 2009.
[3] M. Böhm, G. Koleva, S. Leimeister, C. Riedl, and H. Krcmar. Towards a Generic Value
Network for Cloud Computing. In: Springer LNCS 6296:129-140, 2010.
[4] M. J. Cheon, V. Grover, and J. Teng. Theoretical perspectives on the outsourcing of information
systems. In: Journal of Information Technology 10:209-219, 1995.

12
[5] M. K. Eisenhardt. Agency theory: An assessment and review. In: Academy of Management
Review 14:57-74, 1989.
[6] Forrester Research. Status, Challenges, And Near-Term Tactics For Cloud Services In
Enterprise Outsourcing Deals. November 18, 2009.
[7] I. Foster and S. Tuecke. Describing the elephant: The different faces of IT as service. In: ACM
Queue 3:26-34, 2005.
[8] W.B. Frakes and K. Kang. Software Reuse Research: Status and Future. In: IEEE Transactions
on Software Engineering 31:529-536, 2005.
[9] B. Hayes. Cloud computing. In: Communications of the ACM 51:9-11, 2002.
[10] S. Leimeister, C. Riedl, M. Böhm, and H. Krcmar. The Business Perspective of Cloud
Computing: Actors, Roles, and Value Networks. In: Proceedings of 18th European Conference
on Information Systems (ECIS 2010), Pretoria, South Africa, 2010.
[11] T. Malone, J. Yates, and R. Benjamin. Electronic markets and electronic hierarchies. In:
Communications of the ACM 30:484-497, 1987.
[12] S. Marston, Z. Li, S. Bandyopadhyay, J. Zhang, and A. Ghalsasi. Cloud computing - The
business perspective. In: Decision Support Systems 51:176-189, 2011.
[13] A. B. Mohammed, J. Altmann, and J. Hwang. Cloud Computing Value Chains: Understanding
Businesses and Value Creation in the Cloud. TEMEP Discussion Paper No. 2010:61
[14] O. K. Ngwenyama and N. Bryson. Making the information systems outsourcing decision: A
transaction cost approach to analyzing outsourcing decision problems. In: European Journal of
Operational Research 115:351-361.
[15] M. P. Papazoglou, P. Traverso, S. Dustdar, and F. Leymann. Service-Oriented Computing: State
of the Art and Research Challenges. In: IEEE Computer 40:38-45, 2007.
[16] M. Rappa. The utility business model and the future of computing services. In: IBM Systems
Journal 43:32-42, 2004.
[17] SAP. Cloud Computing and SAP. Where We are And Where We're Going. In: SAPinsider 11,
2010.
[18] L.M. Vaquero, L. Rodero-Merion, J. Caceres, and M. Lindner. A break in the clouds: towards a
cloud definition. In: ACM SIGCOMM Computing Communications Review 39:50-55, 2008.
[19] O. Williamson. Transaction cost economics: The governance of contractual relations. In:
Journal of Law and Economics 22:233-261, 1979.