-
(McKnight, 2007b )
Moreover, new technologies like Cloud computing permanently pose
new challenges to these infrastructures. In order to master the problems arising from this
situation, organizational concepts like the idea of dedicated BI governance structures
(Meredith, 2008)
and BI service management (BISM)
(Horakh, Baars, & Kemper, 2008;
Kashanchi & Toland, 2006)
emerged.
In an organizational context, the Cloud can be regarded as a new outsourcing alternative.
(Baars & Kemper, 2010; Böhm, Leimeister, Riedl, & Krcmar, 2011)
But apart from
wellknown issues with outsourcing, the Cloud comes with additional challenges that have to
be taken into account when integrating this new technology into proprietary BI
architectures.
(Baars & Kemper, 2010; Willem & Jakobus, 2010)
Problems often associated with
Cloud-based BI such as vague data confidentiality, the fear of a vendor lock-in or simply
technological limits indicate that not every service is appropriate to be obtained from the
Cloud.
(Armbrust et al., 2010; Chow et al., 2009; Reyes, 2010)
It is therefore essential to
be clear about the Cloud suitability of individual components and services in order to
include Cloud Computing into a BI architecture successfully. Thus, the goal of this paper
is to gather and structure impacts that have to be considered when planning the use of
Cloud computing in the sophisticated domain of Business Intelligence. This represents
the preliminary work in an on-going larger research process with the ambition to create
an applicable framework for assessing the Cloud suitability of a BI service.
The paper will have the following course. Firstly, basic terms and concepts will be
introduced and discussed on the basis of related literature. This part will mainly cover the
basics of Cloud computing and IT service management as well as their adaption to the
sector of Business Intelligence. Next, the methodical approach used in the research will
be illustrated and the conducted case study will be introduced. Subsequently, the
influences of Cloud BI derived from the reviewed literature and the conducted interviews, will
be discussed in more detail to structure the problem further and carve out challenges.
Afterwards, two real life scenarios that were developed in cooperation with an industrial
partner will illustrate the major issues which mainly originate from the complexity and
the interdependencies in a BI architecture. Finally, the results and next steps will be
summarized in a concluding appraisal.
2
Overview of fundamental concepts and related topics
Since the combination of Cloud computing and BI service management is a rather new
topic, the following section clarifies elementary terms and discusses the basic ideas
behind related concepts and their adaption to the domain of Business Intelligence.
2.1 Cloud computing
There are numerous varying understandings of the term Cloud computing depending on
the point of view.
(Vaquero, Rodero-Merino, Caceres, & Lindner, 2008)
While the Cloud
is often just seen as ”infinite computing resources available on demand”
(Armbrust et al.,
2010, p. 51)
, there are also more sophisticated approaches which constitute Cloud
computing as a new paradigm in IT sourcing that increases agility and enables new business
models.
(Baars & Kemper, 2010; Hayes, 2008; Willem & Jakobus, 2010)
A widely used
definition describes Cloud computing as ”a model for enabling ubiquitous, convenient,
on-demand network access to a shared pool of configurable computing resources (...) that
can be rapidly provisioned and released”
(Mell & Grance, 2009, p. 2)
. The services
offered via the Cloud usually differ in between the supply of fundamental computing
resources (IaaS)
(Bundesverband Informationswirtschaft, 2009; Mell & Grance, 2009;
Rodero-Merino et al., 2010)
up to the provision of complex platform (PaaS)
(Mell &
Grance, 2009; L. Wang et al., 2008; Youseff, Butrico, & Da Silva, 2008)
or application
software (SaaS)
(Bundesverband Informationswirtschaft, 2009; Kern, Lacity, &
Willcocks, 2002; Mell & Grance, 2009)
.
Frequently associated advantages with Cloud computing are cost savings through
UsageBased Pricing models and the possibility to scale resources on demand.
(Armbrust et al.,
2010; Grossman, 2009; Hayes, 2008; Marston, Li, Bandyopadhyay, Zhang, & Ghalsasi,
2011)
It is another common argument that the Cloud simplifies the IT and therefore
improves the level of quality and increases agility.
(Armbrust et al., 2010; Hayes, 2008;
Reeves et al., 2009)
Conversely, the use of Cloud computing also raises significant
challenges such as a lack of Cloud trust and security, insufficient availability, legal issues or
the fear of a vendor lock-in.
(Armbrust et al., 2010; Dillon, Wu, & Chang, 2010; Hayes,
2008; Takabi, Joshi, & Ahn, 2010)
Furthermore, in many cases the integration of a Cloud
solution may not make sense for economic reasons or simply can’t be realized due to
technological limits. (Armbrust et al., 2010; Dillon et al., 2010; Voorsluys, Broberg, &
Buyya, 2011)
2.2 BI services
BI service management (BISM) is an adaption of IT service management (ITSM), which
emerged from the increasing complexity of IT systems in order to implement a more
coherent approach to improve flexibility and manageability of IT environments
(Van Haren
Publishing, 2007; Winniford, Conger, & Erickson-Harris, 2009)
. Therefore, it can be
summarized as “defining, managing, and delivering IT services to support business goals
and customer needs”
(Winniford et al., 2009, p. 153)
, whereat the mentioned service term
represents a combination of technology, people and processes.
(Best Management
Practice, 2012; Van Haren Publishing, 2007)
This definition clarifies the more holistic
organizational character of service management and thus clearly distinguishes it from a
purely technical service-oriented architecture (SOA) approach
(Krafzig, Banke, & Slama,
2005; Papazoglou, 2003)
. Furthermore, service management frameworks used in practice,
like ITIL® or COBIT, usually include additional concepts to design, monitor and
maintain services.
(Unger, 2011; Van Haren Publishing, 2007; Winniford et al., 2009)
BI service management structures and allocates technological and organizational
elements of BI solutions to individual services.
(Kemper, Baars, & Horakh, 2007; Van
Haren Publishing, 2007)
To depict such a BI service the framework shown in figure 1 can
be used. It has been derived in the context of research on BI outsourcing to unbundle the
complex structures of contemporary BI Solutions and slice them into well-defined
services.
(Kemper et al., 2007)
For this purpose, the following three dimensions are defined.
Tool: BI services usually use various subsystems (tools) to fulfill their functions like data
ware houses, analyzing software or front ends. In order to classify these components, it is
useful to refer to existing BI frameworks.
(Kemper et al., 2007)
Business specificity: This dimension depicts the closeness of a service to the business.
The following layers can be distinguished to clarify the specification of this dimension.
(Kemper et al., 2007; Kern et al., 2002)
Hardware: The provision and running of the relevant computing resources such
as storage and other equipment necessary to operate one or more BI components.
Tools: Involves BI software like ETL tools or Data Warehouse software.
Templates: Represent preconfigured applications and prearranged contents that
can be adapted to individual needs (e.g. reference models from software
vendors).
Content: This layer is related to the actual business semantics and therefore
contains concrete business content. An example is the filled instance of a data cube.
Life cycle: The last dimension to be considered is the phase of the application life cycle.
In most cases this dimension can be divided into development and operation since those
are the most significant stages of a service. However, if necessary, this dimension can be
refined to any appropriate granularity.
(Kemper et al., 2007)
To enable a better coverage of customer needs and to achieve a higher scope of
efficiency, compositions of BI services are delivered as holistic BI solutions which contribute to
the user’s business processes.
(Akkermans et al., 2004; Horakh et al., 2008)
The
combination and the multiple usages of BI services lead to interdependencies between particular
components, activities and processes as well as responsibilities of BI solutions.
2.3 Applications of Cloud computing in a Service-based BI environment
This section describes various ways how Cloud computing can be integrated into a
Business Intelligence architecture which mainly differ in their scope and the related
complexity.
(Baars & Kemper, 2010)
Table 1 outlines different scenarios that will be further
described below.
Scenario
Characteristics
Add-on functionality Isolated, short-term, few impacts
Tool integration
Integrated, long-term, high impacts
Solution provision
Complex structures, high impacts
The most straightforward idea is to obtain particular add-on functionality from the Cloud.
These add-ons deliver particular functional blocks for a temporary use.
(Baars & Kemper,
2010; Marston et al., 2011)
Due to its additional and short-term character, this scenario
usually has a low impact on a given BI service environment and is therefore relatively
risk free.
(Baars & Kemper, 2010)
A more long-term oriented approach is the integration or replacement of a tool with a
Cloud-based alternative. This scenario can provide many advantages but depending on
the type of the integrated tool it can also cause far reaching implications for the existing
BI services.
(Baars & Kemper, 2010; Cullen & Willcocks, 2003)
One major issue here,
especially for the domain of BI with its close collaboration between systems and the high
number of dependencies, is the lack of standards for the combination of cloud and
noncloud parts
(Rochwerger et al., 2009; Vaquero et al., 2008)
.
An expansion of this idea is a solution provision where a provider supplies a complete
end to end system in the Cloud. Although this concept is much more complex to integrate
in a proprietary architecture, it might be suited for special purposes e.g. for temporary
systems which have to be set up rapidly.
(Baars & Kemper, 2010)
3
Methodology
Due to the relatively new nature of the subject Cloud computing in the domain of
Business Intelligence, a qualitative design with an exploratory character was used for this
research.
(Stebbins, 2001; Yin, 2010)
The ambition of this work is to examine the impacts of Cloud computing on
Servicebased BI architectures as figure 2 outlines. The study therefore tries to identify and
analyze influencing factors and interdependencies in this context.
In the beginning, a comprehensive literature review
(Auerbach & Silverstein, 2003; Yin,
2010)
was conducted to get an overview of relevant topics and current research subjects.
This review contained a semantic analysis of approximately 100 scientific papers,
whitepapers and market reports mainly from scientific databases like ACM, IEEE Xplore,
SpringerLink and JSTOR as well as from other public accessible sources. Based on this
theoretical foundation, a workshop and two group interviews with business intelligence
experts at Bayer AG, a large German pharmaceutical company, were arranged to get a
practical view on the subject.
The gathered information and an additional analysis of the existing BI infrastructure in
the cooperating company were used to structure challenges and consequences of
Cloudbased Business Intelligence. Since it turned out that the interdependencies of systems and
services play an essential role in the Cloud BI context, a further in-depth interview was
conducted and multiple real life scenarios were developed to examine the related issues
more precisely.
4
Intelligence
Since the usage of Cloud-based services in Business Intelligence can be treated as a
special type of outsourcing
(Baars & Kemper, 2010; Böhm et al., 2011)
, it seems appropriate
to review established IT outsourcing approaches and afterwards carve out the special
challenges with regard to Cloud computing in Business Intelligence services.
In order to justify an IT outsourcing decision, basically two main conditions have to be
met: the technical feasibility has to be given and the gained business value, achieved
through outsourcing, in this particular case over the Cloud, have to predominate the
associated challenges and risks.
(Akomode, Lees, & Irgens, 1998; Böhm et al., 2011; Cullen
& Willcocks, 2003; Loh & Venkatraman, 1995)
Numerous heterogeneous factors and dependencies have to be considered in order to cope
with these questions. Consequently, many established evaluation concepts in this context
use Multi-Criteria Decision Making Methods with hierarchies to prioritize relevant
criteria in order to decrease complexity of the problem.
(Araz, Ozfirat, & Ozkarahan, 2007;
Menzel, Schönherr, Nimis, & Tai, 2011; J.-J. Wang & Yang, 2007)
According to these
methods, the relevant Cloud decision criteria for BI services that were gathered from the
literature review and conducted interviews, can be generalized by the following superior
categories (Figure 3).
Business value: The economic perspective is an essential pillar to assess the Cloud
suitability of a BI service.
(Armbrust et al., 2010; Marston et al., 2011; Mohammed et al.,
2010; Skilton & Director, 2010)
This category summarizes all criteria that affect the
business value in a positive or negative way. Many decision criteria in this class are very
obvious e.g. the reduction of costs or the increasing focus on key competencies. But there
are also criteria which are more complicated to identify, for instance, the enabling of
competitive advantages and new business models. A key problem here is that an
assessment solely based on output measures is of limited value for understanding the whole
range of IT impacts.
(Mooney, Gurbaxani, & Kraemer, 1996)
It is thus essential to
consider the organizational dependencies and the long-range effects on the business value to
constitute a realistic evaluation.
(Mohammed et al., 2010; Mooney et al., 1996)
Technical feasibility: The technical feasibility is a basic condition, required to integrate
the Cloud in an existing environment. Straightforward, the technical suitability of
individual components can be verified by matching hardware and software requirements as
well as technical indicators, such as data transfer volume or availability on Service Level
Agreements of Cloud providers.
(Buyya, Yeo, & Venugopal, 2008; Deb, 2010; Rimal,
Jukan, Katsaros, & Goeleven, 2011)
However, despite this rather intuitive approach, the more profound issue is to assess the
integration complexity of a Cloud-based component in an existing BI architecture. The
sophisticated interdependencies in proprietary BI architectures
(Horakh et al., 2008;
McKnight, 2007b )
and the lack of standards for the technical interoperability of Cloud
services
(Rochwerger et al., 2009; Vaquero et al., 2008)
make integration to one of the
major issues. The complexity manifests itself with factors such as the number of
connected systems or the degree of individualization (e.g. through non-proprietary code).
(Baars
& Kemper, 2010; Deb, 2010)
Risk: Risk can be defined as “the possibility of loss or injury”
(Boehm, 1991, p. 33)
and
plays an important role in any IT outsourcing decision.
(Aubert, Patry, & Rivard, 1998;
Earl, 1996; Hirschheim, Heinzl, & Dibbern, 2009)
There are several approaches to assess
the risk of IT outsourcing. They often consider hidden costs, contractual difficulties,
Service Debasement and the loss of organizational competencies.
(Aubert et al., 1998;
Hirschheim et al., 2009)
In relation to Cloud BI, especially issues in security and trust are frequently referred to as
the main risk factors.
(Armbrust et al., 2010; Voorsluys et al., 2011)
Regulatory and legal
issues such as data privacy or legal liability are also often mentioned in this context.
(Armbrust et al., 2010; Catteddu, 2010; Voorsluys et al., 2011)
These risks mainly
depend on factors like the confidentiality of the handled data or security requirements of the
affected system. Other major concerns are the fear of a vendor lock in, which follows
from the lack of standards and interoperability in Cloud computing
(Marston et al., 2011;
Meredith, 2008)
and other technical risks like insufficient availability or unpredictable
performance
(Armbrust et al., 2010)
. In order to handle these risks the requirements and
the available offers have to be aligned through service level agreements.
(Buyya et al.,
2008; Takabi et al., 2010)
A closer examination of the criteria in these three classes reveals many overlaps and
interdependencies. For instance, a greater technical complexity directly affects the business
value through increasing costs and higher technical and economic risks. Or similarly,
gained business value through a more agile infrastructure strongly depends on the
technical feasibilities of individual components as well as the BI architecture itself.
Complexity of BI services and its influence on the Cloud suitability
Preservative common denominators of the discussed issues are the high complexity of BI
architecture and the strong interdependencies between the involved BI services. This
observation indicates that the complexity of the surrounding BI environment significantly
exerts influence on the Cloud suitability of particular BI services. This complexity mainly
results from highly integrated multi-layer Business Intelligence architectures needed to
provide a holistic management support.
(Baars & Kemper, 2008; McKnight, 2007b )
In
order to implement this enterprise-wide approach, these architectures are interwoven with
all sorts of transactional systems through all hierarchy levels and business functions.
(Chamoni & Gluchowski, 2004; Horakh et al., 2008; Moss & Atre, 2003)
This integrated
structure entails numerous interdependencies between BI solutions and particular BI
services which have to be considered when integrating Cloud-based systems. Furthermore,
the organizational dependencies in such Service-based environments have to be
considered. For instance, new components have to be integrated in existing operating structures
like change or incidence management. This correlation between the complexity of a
system and the ability to use a Cloud approach is also illustrated by the real life scenarios in
the following section.
5
Inspected scenarios
In the course of the conducted case study, several scenarios based upon real life cases
from the Business Intelligence environment of the involved industrial partner were
analyzed. Below, two of these scenarios are used to illustrate the previously discussed issues.
Scenario 1: A temporary Social Media Analytics platform
This case originated from an innovation project with the aim to gather relevant marketing
information from social networks and online communities. The unique character of social
media raised the need for special tools. For instance, data- and text-mining algorithms e.g.
sentiment analysis, were required to gain knowledge from unstructured content.
Furthermore, the large size of data sets raised additional performance challenges.
(Barbier & Liu,
2011; Kaplan & Haenlein, 2010)
The use of a Cloud-based solution in this scenario seems
likely since many providers offer a short-term provision of the needed functionality with
a scalable environment.
Based on the previous classification, this case represents a typical add-on scenario where
special functions are obtained from the Cloud. Due to the inexistence of proprietary tools
and the isolated short-term use in a project, the usage of a Cloud-based solution in this
case would have little impact on the existing BI architecture; hence, there is no need for
an extensive integration process. In addition, the fact that there are barely
interdependencies with existing BI systems entails that the focus of a Cloud evaluation in this case is
mainly on local decision criteria like economic and technical factors of the system.
Scenario 2: Cloud support in a complex best-of-breed BI solution
The second scenario is based upon a large proprietary BI solution. Due to the wide range
of activities, the system over time has evolved into a complex best-of-breed solution;
hence, it contains multiple subsystems across all conceptional levels and is highly
integrated into the Service-based BI architecture as well as into operational systems.
In consideration of the large scale, this system provides a wide range of tool replacement
and add-on possibilities across all conceptional layers. For instance, services on the data
layer can benefit from on demand scaling of computing resources to handle big data
volumes or load peaks. But also more sophisticated applications like a short-term SaaS
provision of particular user interfaces on the access layer are imaginable.
Since the services all interplay in a holistic BI solution, the manifold interdependencies in
this architecture have to be considered when it should be enriched with Cloud-based
components. As previously outlined, trust issues are a major concern when Cloud
providers are involved. Consequently, data confidentiality has to be ensured across all internal
and especially external systems in this end-to-end solution. The strong collaborations
between the various systems in the case complicate this issue further.
A change in the given BI architecture can moreover provoke far-reaching organizational
impacts. It must be guaranteed that services obtained from the Cloud provide the quality
and performance that users and other depending services require. Furthermore, a seamless
integration into the solution life cycle must be possible to ensure a long-term integration.
Therefore, an integration in given service management structures such as change and
incident management is absolutely essential.
Comparison of the scenarios
These two scenarios illustrate the various circumstances and connected problems that
could occur when integrating Cloud approaches into BI architectures.
The obvious difference between the cases is the level of complexity. While scenario one
has an isolated, short-term character and comes with nearly no integration effort. The
second case, in contrast, deals with a large and highly integrated BI solution. The
integration effort in this sophisticated environment is rather large since the manifold
interdependencies between the affected services have to be considered. Moreover, the long-term
replacement of a component in this holistic approach comes with the need of an
organizational integration in existing BI governance structures.
6