Intelligence ArchitectureAuthors

A business intelligence architecture comprises of different unique components to collect, transform, analyze and present the structured and unstructured raw data in simple formats to assist decision makers in making timely decisions. The introduction of service-oriented architecture (SOA) enables reusable services which are accessible over a network on demand. However, there is still a lack of academic literatures on the business intelligence architecture with service-oriented concept. Based upon various references on BI architectures from major vendors, a novel BI architecture that is built on service-oriented concept is presented and described in this paper. The proposed service-oriented architecture enables enterprises to deploy a more agile, flexible, cheaper, reusable, reliable and responsive BI applications in supporting decision making process.


Introduction
Recently, many enterprises are turning their effort towards implementing business intelligence (BI) systems to improve decision making process.A typical business intelligence system include functions such as reporting, multi-dimensional analysis, querying tools, online analytical processing tools, forecasting, data-mining, and advanced visualization capabilities (Microstrategy, 2007).Often, a typical enterprise-level BI architecture constitutes a data warehouse as a common back-end layer.In addition, distributed multi-tier architecture has been widely adopted for modern enterprise BI architecture (Davenport et al., 2007).
According to Erl (2005), service-oriented architecture (SOA) is a group of welldefined services which can be combined, reused and communicated with each other over networks.The introduction of serviceoriented architecture changes many business operations into reusable services which are accessible over a network on demand (IBM, 2007).Within SOA environment, one of the outstanding advantages is individual services that can be accessed without knowing the underlying system platform (Kodali, 2005).Nickull (2005) further states that reusability and repurposing are the main reasons for adopting SOA in implementing enterprise-scale BI systems.SOA enables low-cost system development but with good system quality because it is able to provide a flexible and standardised architecture that supports data sharing and integration of various systems.Other advantages of SOA includes flexibility, responsiveness, reusability, ease of connection, development cost reduction and agibility (IBM Global Technology Services, 2007).
In the traditional approach, client-server architectures are tightly coupled.Any changes made to the client-side might Communications of the IBIMA 2 require relevant changes to be implemented on the server-side as well and vice versa (Cognos Corporation, 2008).However, in the ever evolving business environments, enterprises require an architecture that can easily include, remove or integrate additional services on the fly.In view of this, architectures built based on SOA principles, can separate implementation components from the underlying infrastructure.The separation enables the architecture to support the flexibility, interchangeability, resiliency, security, capacity and availability needed in a modern enterprise BI architecture (IBM Global Technology Services, 2007).
Therefore, this research aims to identify the shortcomings of existing enterprise BI architectures and puts forward a serviceoriented approach.The authors develop a conceptual service-oriented BI architecture model based on the research findings reported in literatures from leading BI vendors such as IBM, Microsoft and Cognos.
In the remaining parts of this paper, the typical business intelligence architecture and its key components are introduced.The next section discusses service-oriented architecture and its benefits.Lastly, we study existing modern BI architectures and propose a new conceptual model on service-oriented business intelligence architecture.

Literature Review
According to Ponniah (2001), architecture is the combination of different unique components which provide a means to store data, and deliver information to users.Architecture is a template that contains rules and functions used for serving the business requirements.The various elements in architecture such as the standards, measurements, designs and other supporting techniques, aim to enable the smooth data flow from the source to the destination within a framework (Ponniah, 2001).
A BI architecture that comprises of different systems, applications, and processes in an enterprise enables business decision makers in accessing the valuable information for complicated analytical processes.The BI architecture should be able to provide the correct information promptly in supporting the decision making processes.Hence, the information must be distributed via multiple channels including emails and page alerts spreadsheets, analytic queries, scorecards and dashboards (Davenport et al., 2007).In other words, BI architecture is anything that enables the transformation of data into useful information in decisionmaking process and to acquire evolving business advantages.

According
to Howson (2007), a fundamental architecture is the essential elemental for all BI system deployments.On one hand, the lower level BI architecture can be formed by the transactional systems and front-end tools.On the other hand, a higher level BI architecture deployment may consist of data marts, data warehouses, ETL (extract, transform and load) tools and BI front-end tools.
A study was separately performed on the BI architectures by two major BI vendors, namely, IBM and Microsoft.The authors found that both of the proposed architectures consist of five similar layers: data access/presentation layer, data analysis layer, data repository/data storage layer, data integration layer, and data source layer.• A presentation tier that handles interaction with users over the network; • An application tier to handle all BI processing with special-built services; and • A data tier that provides access various data sources.

The Conceptual Model of SOA for Enterprise Business Intelligence
One of the main reasons for a BI project failure is largely due to the selection of inappropriate BI tools that fails to meet the specified business requirements.Therefore, Ponniah (2001) recommends that enterprises should design the architecture first, only then select the tools to match the functions and services stipulated for the architectural components.According to Friedman et al. (2004), there is a limited awareness of the value of architecture in the BI context.This creates challenges in data quality for many enterprises and the "hidden" aspects such as reliability, scalability and flexibility of BI are not being emphasized.According to Friedman et al. (2004), there is a limited awareness of the value of architecture in the BI context.This creates challenges in data quality for many enterprises and the "hidden" aspects such as reliability, scalability and flexibility of BI are not being emphasized.2007) assert that developing a robust BI system is more than just gathering and storing huge amount of data since it involves facets such as data quality, business processes, incentives, skills, organizational cultures, and sponsorships.Therefore, a BI architecture should be considered first when scoping a BI solution so that a more appropriate BI solution can be developed in meeting the actual needs of an enterprise.The finding was supported by Friedman et al. (2004), who state that architecture is the foundation of BI and paying attention to the architecture will ensure success in BI implementation.Watson et al. (2005) also emphasize that architecture has significant impact when users can easily access the business data, which in turn leads to improved decisionmaking capabilities.In order to meet the requirements of an emerging business environment, a BI architecture should be designed in a flexible and adaptable manner (Davenport et al. 2007).Invariably, a rigid BI architecture design will be better off serving the needs of an ever-changing business environment.

Davenport el al. (
In view of the challenges and limited capabilities of existing SOA-based architecture, a six-tier SOA-based BI architecture conceptual model (Figure 1) is proposed to deliver BI capabilities in the constantly evolving business environments.The details of the model's tiers are described in the following sections.

Fig1. Proposed Conceptual Model
Communications of the IBIMA 6

Presentation Tier
The top layer is the data presentation or data access layer which provides access to the users via various means such as host computers, portal, web service or even handheld mobile devices.Furthermore, users are able to view reports in various formats such as chart, graph, table, report, dashboard, scorecard, metrics, key performance indicator (KPI).The reports are equipped with navigation capabilities like drilling, pivoting, drag-drop so that users can seamlessly analyze the data in any reports.

Communication Tier
Open interface is used by all services to communicate with each others and it should take place in the communication tier.This can achieve location transparency among services.Open web standards, such as Simple Object Access Protocol (SOAP), Extensible Markup Language (XML) and Web Services Definition Language (WSDL) should be adopted for communication purpose.

Analytic Tier
Since a BI system contains huge volume of data, tools and techniques from data integration layer, data analysis layer is critical for managing, summarizing, querying and analyzing the data.It provides analytics functions such as ad-hoc query, reporting, modelling, embedded analytics, Online Analytical Processing (OLAP), data mining and collaboration.Data is organized into multidimensional cubes based on a dimensional model which enables summarizing information for analytical queries in an efficient manner.Data analysis leverages the information stored in data storage.

Processed Repository Tier
Data Storage is the end result of Data Source and Data Integration layers.Data Storage can be termed as Data Warehouse (DW)/Data Mart (DM).According to Microsoft architecture, it involves elements such as dimensional modelling, partitions and indexes.Meanwhile IBM proposes operational data stores, staging areas and metadata in this layer.

Data Integration Tier
BI often involves analysis of aggregated and integrated data from various operational systems.Data is extracted, cleaned and filtered from the operational systems before it is loaded into a data warehouse.A data integration layer primarily consists of ETL (extracting, transforming, and loading) process, whilst it also involves processes such as profiling, cleansing, and staging (Howson, 2007).

Data Source Tier
The most fundamental data collection is done at the data source layer (bottom layer) where it includes different sources environment, and systems including different platforms and operating systems).Data are collected from various resources.It could be operational data which are generated by internal departments, unstructured data, informational data and external data.Operational systems, which serve as the first entry for quantitative data, can be sourced from any transaction processing systems, enterprise resource planning systems or any other systems.Operational data can be generated from internal enterprise operational systems such as production system, sales and marketing system, supply chain system, and accounting system.Data may also come from external sources such as from suppliers, customers or any information on the Internet.The issues of confidentiality, integrity, and authenticity are the most important concerns in any computer system.Data communication over pervasive network increases the vulnerability to computer security risks like viruses and intrusion.Some of the basic computer security practices like firewalls, routers, encryption and authentication, thereby, help to protect and prevent BI systems from unauthorized access.Therefore, it is proposed to implement the mentioned security measures above in two tiers.First, a security tier between data source tier and data integration tier.Second, a security tier between communication tier and presentation tier.Data sources contain data from various resources.Therefore, it is extremely important to make sure that the input data is free of viruses and intruders.Before the information is presented to the presentation tier, it is proposed to allocate some security measures to clean the reports which will be delivered to any digital devices in various formats (Fenzel and Fenzel, 2004).

Conclusion
This paper presents a service-oriented architecture for enterprise-scale business intelligence environment.Based on reference BI architectures from major BI vendors, a modern BI architecture that is built on service-oriented concept is proposed and described in this paper.An effective service-oriented BI architecture enables enterprises to deploy more flexible and reliable BI applications to a greater number of business stakeholders in improving the decision support.With the proposed architecture, enterprises can incorporate BI elements without having to reengineer their existing IT infrastructure and associated complexities in systems administration, training, and server management.

Table 1 : Comparison of IBM (2004) and Microsoft (2006) BI Architectures
Table 1 presents the result of comparison between IBM's BI Reference Architecture and Microsoft Conceptual System Architecture in accordance to layers.

Table 2 .
Communications of the IBIMA 4